Re: Neural mechanisms of octave equivalence ("Richard F. Lyon" )


Subject: Re: Neural mechanisms of octave equivalence
From:    "Richard F. Lyon"  <dicklyon@xxxxxxxx>
Date:    Sat, 24 Sep 2016 10:29:41 -0700
List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>

--047d7b604aac829177053d443dc4 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Alain, right, I didn't mention harmonic templates; yes, that mechanism can work, but if birds mostly use narrowband harmonic-poor songs, and don't develop such templates, but do match on individual peaks, then no octave equivalence would be expected. Personally, I don't think the harmonic template approach is plausible. It's too hard to see how to make it work with plausible mechanisms that follow what the cochlea does, in either birds or mammals. Dick On Sat, Sep 24, 2016 at 10:17 AM, Alain de Cheveigne < alain.de.cheveigne@xxxxxxxx> wrote: > Hi Dick, > > The mechnanisms I mentioned don=E2=80=99t depend on the tones having harm= onics. > This is easy to see for autocorrelation: the property we invoke is that > peaks for the lower tone coincide with peaks of the octave tone, which is > true whether or not the tones have harmonics. The same holds for pattern > matching: the fundamental and and harmonics of the octave tone match the > harmonic template defined by the lower tone (whether that tone contains > harmonics or not). The harmonic template is an abstract pattern that can > be matched by a harmonic spectrum but also by a single component. > > Of course it could be that we (or birds) do perceive octave similarity by > matching patterns formed by the harmonics of the tones. That would be ye= t > another model. > > Alain > > > > > On 24 Sep 2016, at 18:39, Richard F. Lyon <dicklyon@xxxxxxxx> wrote: > > > > I agree with Alain, but looked at the paper and have a few more comment= s: > > https://www.researchgate.net/profile/Lauren_Guillette/ > publication/235376278_Chickadees_fail_standardized_ > operant_tests_for_octave_equivalence/links/09e4151228c5ace7d7000000.pdf > > > > Octave equivalence is pretty strong for tones with enough harmonics, fo= r > reasons that Alain describes. This paper shows that humans have some > octave generalization even with pure sine waves, and that the birds do > not. This likely points to different mechanisms. > > > > For matching based on common frequencies of partials, you need some > partials in common, which is not the case here, due to the signals being > sine waves (no upper partials). For matching based on common periods, or > common peaks in autocorrelation functions, sine waves an octave apart are > close, because the higher one is also periodic at the period of the lower > one. > > > > So maybe this argues that humans use more period-based matching and > birds use only frequency (cochlear place) matching? Maybe the experiment > should be repeated with tones that have at least a second harmonic, and s= ee > if that leads to birds doing octave generalization by matching one tone's > fundamental to another's second harmonic? This would be a better way to > get at pitch height versus chroma, perhaps. > > > > Dick > > > > > > On Sat, Sep 24, 2016 at 12:59 AM, Alain de Cheveigne < > alain.de.cheveigne@xxxxxxxx> wrote: > > Hi Ani, > > > > Octave =E2=80=9Cequivalence=E2=80=9D is an emergent property of both pa= ttern-matching > and autocorrelation models of pitch. All harmonics of the tone at the > octave belong to the harmonic series of the lower tone. Likewise > autocorrelation peaks of the lower tone coincide with peaks of the tone a= t > the octave. Some neural instantiations of these models are Shihab Shamma= =E2=80=99s > harmonic template model, or Cariani=E2=80=99s work on autocorrelation (ba= sed on > Licklider=E2=80=99s ideas), and there are many others. Whether or not an= y specific > model is supported by anatomical or electrophysiological data is less cle= ar. > > > > Actually =E2=80=9Cequivalence=E2=80=9D is a misnomer. The relation is n= ot commutative: > the harmonics of the lower tone do not all belong to the harmonic series = of > the octave. Likewise peaks of the autocorrelation of the octave tone are > not all peaks of the lower tone. Thus these models would predict an > asymmetry in the perceptual similarity between octaves (i.e. an octave to= ne > =E2=80=9Cresembles=E2=80=9D the lower tone but not vice-versa). I don=E2= =80=99t know of any > relevant behavioral data or music-theoretical results on this. > > > > Alain > > > > =E2=80=94 > > de Cheveign=C3=A9, A. (2005) Pitch perception models. In: Pitch - Neura= l > coding and perception (Plack C, Oxenham A, eds). New York: Springer, > 169-233. (http://audition.ens.fr/adc/pdf/2005_pitch_SHAR.pdf) > > Shamma S, and Klein D (2000) The case of the missing pitch templates: > how harmonic templates emerge in the early auditory system. J Acoust Soc = Am > 107:2631-2644. > > Cariani PA, and Delgutte B (1996b) Neural correlates of the pitch of > complex tones. II. Pitch shift, pitch ambiguity, phase-invariance, pitch > circularity, rate-pitch and the dominance region for pitch. J Neurophysio= l > 76:1717-1734. > > Licklider JCR (1951) A duplex theory of pitch perception (reproduced in > Schubert 1979, 155-160). Experientia 7:128-134. > > > > > > > On 23 Sep 2016, at 13:06, Patel, Aniruddh D. <a.patel@xxxxxxxx> > wrote: > > > > > > Dear List, > > > > > > Is anyone aware on theoretical or empirical papers on the neural > mechanisms of octave equivalence in auditory perception? > > > > > > Interestingly, recent works suggests that songbirds may not perceive > octave equivalence: > > > > > > Hoeschele, M., Weisman, R. G., Guillette, L. M., Hahn, A. H., & > Sturdy, C. B. (2013). Chickadees fail standardized operant tests for octa= ve > equivalence. Animal cognition, 16(4), 599-609. > > > > > > Thanks, > > > > > > Ani Patel > > > > > > Aniruddh D. Patel > > > Professor > > > Dept. of Psychology > > > Tufts University > > > 490 Boston Ave. > > > Medford, MA 02155 > > > > > > Senior Fellow > > > Canadian Institute for Advanced Research (CIFAR) > > > Azrieli Program in Brain, Mind, & Consciousness > > > > > > a.patel@xxxxxxxx > > > http://ase.tufts.edu/psychology/people/patel/ > > > > --047d7b604aac829177053d443dc4 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable <div dir=3D"ltr"><div>Alain, right, I didn&#39;t mention harmonic templates= ; yes, that mechanism can work, but if birds mostly use narrowband harmonic= -poor songs, and don&#39;t develop such templates, but do match on individu= al peaks, then no octave equivalence would be expected.<br><br></div><div>P= ersonally, I don&#39;t think the harmonic template approach is plausible. I= t&#39;s too hard to see how to make it work with plausible mechanisms that = follow what the cochlea does, in either birds or mammals.<br><br></div>Dick= <br><br></div><div class=3D"gmail_extra"><br><div class=3D"gmail_quote">On = Sat, Sep 24, 2016 at 10:17 AM, Alain de Cheveigne <span dir=3D"ltr">&lt;<a = href=3D"mailto:alain.de.cheveigne@xxxxxxxx" target=3D"_blank">alain.de.chevei= gne@xxxxxxxx</a>&gt;</span> wrote:<br><blockquote class=3D"gmail_quote" style= =3D"margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi Dick,= <br> <br> The mechnanisms I mentioned don=E2=80=99t depend on the tones having harmon= ics.=C2=A0 This is easy to see for autocorrelation: the property we invoke = is that peaks for the lower tone coincide with peaks of the octave tone, wh= ich is true whether or not the tones have harmonics.=C2=A0 The same holds f= or pattern matching: the fundamental and and harmonics of the octave tone m= atch the harmonic template defined by the lower tone (whether that tone con= tains harmonics or not).=C2=A0 The harmonic template is an abstract pattern= that can be matched by a harmonic spectrum but also by a single component.= <br> <br> Of course it could be that we (or birds) do perceive octave similarity by m= atching patterns formed by the harmonics of the tones.=C2=A0 That would be = yet another model.<br> <span class=3D"HOEnZb"><font color=3D"#888888"><br> Alain<br> </font></span><div class=3D"HOEnZb"><div class=3D"h5"><br> <br> <br> &gt; On 24 Sep 2016, at 18:39, Richard F. Lyon &lt;<a href=3D"mailto:dickly= on@xxxxxxxx">dicklyon@xxxxxxxx</a>&gt; wrote:<br> &gt;<br> &gt; I agree with Alain, but looked at the paper and have a few more commen= ts:<br> &gt; <a href=3D"https://www.researchgate.net/profile/Lauren_Guillette/publi= cation/235376278_Chickadees_fail_standardized_operant_tests_for_octave_equi= valence/links/09e4151228c5ace7d7000000.pdf" rel=3D"noreferrer" target=3D"_b= lank">https://www.researchgate.net/<wbr>profile/Lauren_Guillette/<wbr>publi= cation/235376278_<wbr>Chickadees_fail_standardized_<wbr>operant_tests_for_o= ctave_<wbr>equivalence/links/<wbr>09e4151228c5ace7d7000000.pdf</a><br> &gt;<br> &gt; Octave equivalence is pretty strong for tones with enough harmonics, f= or reasons that Alain describes.=C2=A0 This paper shows that humans have so= me octave generalization even with pure sine waves, and that the birds do n= ot.=C2=A0 This likely points to different mechanisms.<br> &gt;<br> &gt; For matching based on common frequencies of partials, you need some pa= rtials in common, which is not the case here, due to the signals being sine= waves (no upper partials).=C2=A0 For matching based on common periods, or = common peaks in autocorrelation functions, sine waves an octave apart are c= lose, because the higher one is also periodic at the period of the lower on= e.<br> &gt;<br> &gt; So maybe this argues that humans use more period-based matching and bi= rds use only frequency (cochlear place) matching?=C2=A0 Maybe the experimen= t should be repeated with tones that have at least a second harmonic, and s= ee if that leads to birds doing octave generalization by matching one tone&= #39;s fundamental to another&#39;s second harmonic?=C2=A0 This would be a b= etter way to get at pitch height versus chroma, perhaps.<br> &gt;<br> &gt; Dick<br> &gt;<br> &gt;<br> &gt; On Sat, Sep 24, 2016 at 12:59 AM, Alain de Cheveigne &lt;<a href=3D"ma= ilto:alain.de.cheveigne@xxxxxxxx">alain.de.cheveigne@xxxxxxxx</a>&gt; wrote:<br= > &gt; Hi Ani,<br> &gt;<br> &gt; Octave =E2=80=9Cequivalence=E2=80=9D is an emergent property of both p= attern-matching and autocorrelation models of pitch. All harmonics of the t= one at the octave belong to the harmonic series of the lower tone.=C2=A0 Li= kewise autocorrelation peaks of the lower tone coincide with peaks of the t= one at the octave.=C2=A0 Some neural instantiations of these models are Shi= hab Shamma=E2=80=99s harmonic template model, or Cariani=E2=80=99s work on = autocorrelation (based on Licklider=E2=80=99s ideas), and there are many ot= hers.=C2=A0 Whether or not any specific model is supported by anatomical or= electrophysiological data is less clear.<br> &gt;<br> &gt; Actually =E2=80=9Cequivalence=E2=80=9D is a misnomer. The relation is = not commutative: the harmonics of the lower tone do not all belong to the h= armonic series of the octave.=C2=A0 Likewise peaks of the autocorrelation o= f the octave tone are not all peaks of the lower tone.=C2=A0 Thus these mod= els would predict an asymmetry in the perceptual similarity between octaves= (i.e. an octave tone =E2=80=9Cresembles=E2=80=9D the lower tone but not vi= ce-versa).=C2=A0 I don=E2=80=99t know of any relevant behavioral data or mu= sic-theoretical results on this.<br> &gt;<br> &gt; Alain<br> &gt;<br> &gt; =E2=80=94<br> &gt; de Cheveign=C3=A9, A. (2005) Pitch perception models. In: Pitch - Neur= al coding and perception (Plack C, Oxenham A, eds). New York: Springer, 169= -233. (<a href=3D"http://audition.ens.fr/adc/pdf/2005_pitch_SHAR.pdf" rel= =3D"noreferrer" target=3D"_blank">http://audition.ens.fr/adc/<wbr>pdf/2005_= pitch_SHAR.pdf</a>)<br> &gt; Shamma S, and Klein D (2000) The case of the missing pitch templates: = how harmonic templates emerge in the early auditory system. J Acoust Soc Am= 107:2631-2644.<br> &gt; Cariani PA, and Delgutte B (1996b) Neural correlates of the pitch of c= omplex tones. II. Pitch shift, pitch ambiguity, phase-invariance, pitch cir= cularity, rate-pitch and the dominance region for pitch. J Neurophysiol 76:= 1717-1734.<br> &gt; Licklider JCR (1951) A duplex theory of pitch perception (reproduced i= n Schubert 1979, 155-160). Experientia 7:128-134.<br> &gt;<br> &gt;<br> &gt; &gt; On 23 Sep 2016, at 13:06, Patel, Aniruddh D. &lt;<a href=3D"mailt= o:a.patel@xxxxxxxx">a.patel@xxxxxxxx</a>&gt; wrote:<br> &gt; &gt;<br> &gt; &gt; Dear List,<br> &gt; &gt;<br> &gt; &gt; Is anyone aware on theoretical or empirical papers on the neural = mechanisms of octave equivalence in auditory perception?<br> &gt; &gt;<br> &gt; &gt; Interestingly, recent works suggests that songbirds may not perce= ive octave equivalence:<br> &gt; &gt;<br> &gt; &gt; Hoeschele, M., Weisman, R. G., Guillette, L. M., Hahn, A. H., &am= p; Sturdy, C. B. (2013). Chickadees fail standardized operant tests for oct= ave equivalence. Animal cognition, 16(4), 599-609.<br> &gt; &gt;<br> &gt; &gt; Thanks,<br> &gt; &gt;<br> &gt; &gt; Ani Patel<br> &gt; &gt;<br> &gt; &gt; Aniruddh D. Patel<br> &gt; &gt; Professor<br> &gt; &gt; Dept. of Psychology<br> &gt; &gt; Tufts University<br> &gt; &gt; 490 Boston Ave.<br> &gt; &gt; Medford, MA 02155<br> &gt; &gt;<br> &gt; &gt; Senior Fellow<br> &gt; &gt; Canadian Institute for Advanced Research (CIFAR)<br> &gt; &gt; Azrieli Program in Brain, Mind, &amp; Consciousness<br> &gt; &gt;<br> &gt; &gt; <a href=3D"mailto:a.patel@xxxxxxxx">a.patel@xxxxxxxx</a><br> &gt; &gt; <a href=3D"http://ase.tufts.edu/psychology/people/patel/" rel=3D"= noreferrer" target=3D"_blank">http://ase.tufts.edu/<wbr>psychology/people/p= atel/</a><br> &gt;<br> <br> </div></div></blockquote></div><br></div> --047d7b604aac829177053d443dc4--


This message came from the mail archive
/var/www/html/postings/2016/
maintained by:
DAn Ellis <dpwe@ee.columbia.edu>
Electrical Engineering Dept., Columbia University