Re: information transmission analyses ("Iftikhar Riaz (Lance) Nizami" )


Subject: Re: information transmission analyses
From:    "Iftikhar Riaz (Lance) Nizami"  <Nizamii2@xxxxxxxx>
Date:    Fri, 27 Feb 2009 16:52:23 EST
List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>

-------------------------------1235771543 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit There is little remaining debate about whether information-theoretical analysis in psychology led anywhere besides "down the garden path". The people doing that research all thought they had a "powerful tool" and said so many times over in plain English. In fact, all they had was a statistic that, like all statistics, tells you nothing about the thing that you actually want to understand, i.e. the underlying physical mechanisms. The psychologists' information statistic did not even indicate the thing that people imagined that it did (i.e. transmitted information); that relation was always assumed, but never proven. Whether the fad of applying information theory to neurons (which is hardly new) will lead anywhere in particular remains to be seen. Like the perceptual psychologists, the neuroscience users may be making unjustified tacit assumptions. There are plenty of "users' but few "understanders" out there. Application to neurons might seem to indicate that one method of encoding appears to be more "efficient" (say) than another, but it can never, ever tell you how the system actually operates. Never. The psychologists still don't understand this point ... - Lance Nizami In a message dated 2/27/2009 4:14:56 P.M. Eastern Standard Time, paris@xxxxxxxx writes: Thanks for the interesting bibliography. One note I would add though, is that, with the exception of the Luce paper, these publications precede a rather significant boom in the use of info theory in the computational neuroscience world. The use of a communication channel analog in cognitive psychology might be a stretch, but for neural- level perception studies it has become quite a powerful tool. Seeing how these studies are slowly branching towards high-level perception I would say that the debate (if any) is probably still open. Best, Paris On Feb 27, 2009, at 1:04 PM, Iftikhar Riaz (Lance) Nizami wrote: > Dear Dr. Pinter and others, > > A comprehensive critical review that argues that "The human being is > NOT a Shannon communication channel" does not exist in print, to my > knowledge. And no editor is going to allow such a paper to be > published, because there are too many prominent people who stand to > lose if their misinterpretations are made public. Some of those > people are senior editors of well-known journals. There are, > however, a number of published papers that question whether the use > of information-transmission measures in psychology was really useful: > > Cronbach, L.J. (1955) On the non-rational application of > information measures in psychology. In: Information theory in > psychology (pp. 14-30), ed. H. Quastler. Glencoe, Illinois: The Free > Press. > Luce, R.D. (2003) Whatever happened to information theory in > psychology? Rev. Gen. Psych. 7, 183-188. > Gregory, R.L. (1980) Whatever happened to information theory? (2) > Perception, 9, 489-492. > Laming, D. (1973) Mathematical Psychology. Academic Press, NY. > Ashby, F.G. (1995). Resurrecting information theory. The American > Journal of Psychology, 108, 609-614. > > A outstanding (and unfairly ignored) paper that deals decisively > with the issue of what "informational" absolute judgment experiments > actually measure is: > > Siegel, W. (1972) Memory effects in the method of absolute > judgment. Journal of Experimental Psychology, 94, 121-131. > > - Lance Nizami PhD, Decatur, GA 30030 > > > > In a message dated 2/26/2009 11:30:22 P.M. Eastern Standard Time, andzsinszan@xxxxxxxx > writes: > Dear Lance Nizami, > > What you wrote regarding Information Theory is quite interesting to > me, as I daily meet people doing automatic speech recognition, whom > are really difficulty to convince that there is life beyond > information theory and Markov chains. > Can you please give us some references that argues that "The human > being is NOT a Shannon communication channel". > Or at least focusing on the controversial relation between perception > and information theory. > > I appreciate your help. > > Best regards. > > Gabor Pinter > > > > > > > > On Wed, Feb 25, 2009 at 1:44 AM, Iftikhar Riaz (Lance) Nizami > <Nizamii2@xxxxxxxx> wrote: > > I did my Master's on this subject, and had to chance to read > Attneave and > > the rest of the literature in depth. Information theory as > applied to > > psychology, as popularized by Attneave after Garner & Hake, has > nothing at > > all to do with transmission of anything. When applied to > traditional > > absolute judgment (identification) experiments, for example, it is > merely an > > alternative measure of short-term sensory memory. Which has > nothing to do > > with Claude Shannon's "general communications system". In your > case, memory > > capacity is probably what the analysis will indicate. I am > surprised that > > anyone uses the Garner-Hake information approach anymore; > mathematical > > psychologists (Duncan Luce, Donald Laming, Sandy MacRae, etc.) > recognized > > its severe limitations years ago and abandoned it. So > (eventually) did > > acoustics users like Neff and Lutfi. The human being is NOT a > > Shannon communication channel. Granted, the measures that result > from > > processing the confusion matrix are covariance measures of a > sort. In that > > case, ordinary covariance measures may provide a more meaningful > way of > > analyzing your data. In response to your question, then, none of > TRANS, > > TRANS/INPUT or TRANS/TI are appropriate measures for your purposes. > > Information theory doesn't work that way. Attneave got the math > right, > > but that had already been done; but Attneave's interpretations > (after Garner > > & Hake) were all dead wrong. - Lance Nizami BSc (Physics) MSc > (Biomedical > > Engineering) PhD (Psychophysics), Decatur, GA 30030 > > > > In a message dated 2/24/2009 4:54:12 A.M. Eastern Standard Time, > > caro_jacquier@xxxxxxxx writes: > > > > Dear members, > > > > I would like to analyse phonetic feature transmission (especially > voicing > > and place of articulation of french plosive consonants) on the > basis of > > individual confusion matrices. My study is about compressed speech > > perception in dyslexic adults. > > I have already downloaded the software for information transfer > and SINFA > > analysis (Stuart Rosen ) and I have run the analyses (on 32 > subjects). > > My first question is: which value is the more relevant in the > report (TRANS, > > TRANS/INPUT or TRANS/TI) ? > > My second question is about individual confusion matrices: Do I > have to do > > one analyse per subject ? > > > > I hope that someone could help me. > > Thank you very much. > > Best regards, > > > > Caroline > > > > > > > > ________________________________ > > A Good Credit Score is 700 or Above. See yours in just 2 easy steps! > > A Good Credit Score is 700 or Above. See yours in just 2 easy steps! **************A Good Credit Score is 700 or Above. See yours in just 2 easy steps! (http://pr.atwola.com/promoclk/100126575x1218822736x1201267884/aol?redir=http:%2F%2Fwww.freecreditreport.com%2Fpm%2Fdefault.aspx%3Fsc%3D668072%26hmpgID %3D62%26bcd%3DfebemailfooterNO62) -------------------------------1235771543 Content-Type: text/html; charset="US-ASCII" Content-Transfer-Encoding: quoted-printable <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN"> <HTML><HEAD> <META http-equiv=3DContent-Type content=3D"text/html; charset=3DUS-ASCII"> <META content=3D"MSHTML 6.00.6000.16809" name=3DGENERATOR></HEAD> <BODY id=3Drole_body style=3D"FONT-SIZE: 10pt; COLOR: #000000; FONT-FAMILY:=20= Arial"=20 bottomMargin=3D7 leftMargin=3D7 topMargin=3D7 rightMargin=3D7><FONT id=3Drol= e_document=20 face=3DArial color=3D#000000 size=3D2> <DIV>There is little remaining debate about whether information-theoretical=20 analysis in psychology led anywhere besides "down the garden path".&nbsp; Th= e=20 people doing that research all thought they had a "powerful tool" and said s= o=20 many times over in plain English.&nbsp; In fact, all they had was a=20 <EM>statistic</EM> that, like all statistics, tells you nothing about the th= ing=20 that you actually want to understand, i.e. the underlying physical=20 mechanisms.</DIV> <DIV>&nbsp;</DIV> <DIV>The psychologists' information statistic did not even indicate the thin= g=20 that people imagined that it did (i.e. transmitted information); that relati= on=20 was always assumed, but never proven.&nbsp; Whether the fad of applying=20 information theory to neurons (which is hardly new) will lead anywhere in=20 particular remains to be seen.&nbsp; Like the perceptual psychologists, the=20 neuroscience users may be making&nbsp;unjustified tacit=20 assumptions.&nbsp;&nbsp;There are plenty of "users' but few "understanders"=20= out=20 there.&nbsp; Application to neurons&nbsp;might seem to&nbsp;indicate that on= e=20 method of encoding appears to be more "efficient" (say) than another, but it= can=20 never, ever tell&nbsp;you how the system actually operates.&nbsp; Never.&nbs= p;=20 The psychologists still don't understand this point ... - Lance Nizami</DIV> <DIV>&nbsp;</DIV> <DIV> <DIV>In a message dated 2/27/2009 4:14:56 P.M. Eastern Standard Time,=20 paris@xxxxxxxx writes:</DIV> <BLOCKQUOTE=20 style=3D"PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: blue 2px solid"><= FONT=20 style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size= =3D2>Thanks=20 for the interesting bibliography.&nbsp; One note I would add though,&nbsp;= =20 <BR>is that, with the exception of the Luce paper, these publications&nbsp= ;=20 <BR>precede a rather significant boom in the use of info theory in the&nbs= p;=20 <BR>computational neuroscience world.&nbsp; The use of a communication=20 channel&nbsp; <BR>analog in cognitive psychology might be a stretch, but f= or=20 neural- <BR>level perception studies it has become quite a powerful=20 tool.&nbsp; Seeing&nbsp; <BR>how these studies are slowly branching toward= s=20 high-level perception I&nbsp; <BR>would say that the debate (if any) is=20 probably still open.<BR><BR>Best,<BR>Paris<BR><BR><BR>On Feb 27, 2009, at=20= 1:04=20 PM, Iftikhar Riaz (Lance) Nizami wrote:<BR><BR>&gt; Dear Dr. Pinter and=20 others,<BR>&gt;<BR>&gt; A comprehensive critical review that argues that "= The=20 human being is&nbsp; <BR>&gt; NOT a Shannon communication channel" does no= t=20 exist in print, to my&nbsp; <BR>&gt; knowledge.&nbsp; And no editor is goi= ng=20 to allow such a paper to be&nbsp; <BR>&gt; published, because there are to= o=20 many prominent people who stand to&nbsp; <BR>&gt; lose if their=20 misinterpretations are made public.&nbsp; Some of those&nbsp; <BR>&gt; peo= ple=20 are senior editors of well-known journals.&nbsp; There are,&nbsp; <BR>&gt;= =20 however, a number of published papers that question whether the use&nbsp;=20 <BR>&gt; of information-transmission measures in psychology was really=20 useful:<BR>&gt;<BR>&gt; Cronbach, L.J. (1955)&nbsp; On the non-rational=20 application of&nbsp; <BR>&gt; information measures in psychology.&nbsp; In= :=20 Information theory in&nbsp; <BR>&gt; psychology (pp. 14-30), ed. H. Quastl= er.=20 Glencoe, Illinois: The Free&nbsp; <BR>&gt; Press.<BR>&gt; Luce, R.D. (2003= )=20 Whatever happened to information theory in&nbsp; <BR>&gt; psychology? Rev.= =20 Gen. Psych. 7, 183-188.<BR>&gt; Gregory, R.L. (1980)&nbsp; Whatever happen= ed=20 to information theory? (2)&nbsp;&nbsp; <BR>&gt; Perception, 9,=20 489-492.<BR>&gt; Laming, D. (1973) Mathematical Psychology. Academic Press= ,=20 NY.<BR>&gt; Ashby, F.G. (1995). Resurrecting information theory. The=20 American&nbsp; <BR>&gt; Journal of Psychology, 108, 609-614.<BR>&gt;<BR>&g= t; A=20 outstanding (and unfairly ignored) paper that deals decisively&nbsp; <BR>&= gt;=20 with the issue of what "informational" absolute judgment experiments&nbsp;= =20 <BR>&gt; actually measure is:<BR>&gt;<BR>&gt; Siegel, W. (1972)&nbsp; Memo= ry=20 effects in the method of absolute&nbsp; <BR>&gt; judgment.&nbsp; Journal o= f=20 Experimental Psychology, 94, 121-131.<BR>&gt;<BR>&gt; - Lance Nizami PhD,=20 Decatur, GA 30030<BR>&gt;<BR>&gt;<BR>&gt;<BR>&gt; In a message dated 2/26/= 2009=20 11:30:22 P.M. Eastern Standard Time, andzsinszan@xxxxxxxx <BR>&gt;&nbsp;=20 writes:<BR>&gt; Dear Lance Nizami,<BR>&gt;<BR>&gt; What you wrote regardin= g=20 Information Theory is quite interesting to<BR>&gt; me, as I daily meet peo= ple=20 doing automatic speech recognition, whom<BR>&gt; are really difficulty to=20 convince that there is life beyond<BR>&gt; information theory and Markov=20 chains.<BR>&gt; Can you please give us some references that argues that "T= he=20 human<BR>&gt; being is NOT a Shannon communication channel".<BR>&gt; Or at= =20 least focusing on the controversial relation between perception<BR>&gt; an= d=20 information theory.<BR>&gt;<BR>&gt; I appreciate your help.<BR>&gt;<BR>&gt= ;=20 Best regards.<BR>&gt;<BR>&gt; Gabor=20 Pinter<BR>&gt;<BR>&gt;<BR>&gt;<BR>&gt;<BR>&gt;<BR>&gt;<BR>&gt;<BR>&gt; On=20= Wed,=20 Feb 25, 2009 at 1:44 AM, Iftikhar Riaz (Lance) Nizami<BR>&gt;=20 &lt;Nizamii2@xxxxxxxx&gt; wrote:<BR>&gt; &gt; I did my Master's on this=20 subject, and had to chance to read&nbsp; <BR>&gt; Attneave and<BR>&gt; &gt= ;=20 the rest of the literature in depth.&nbsp; Information theory as&nbsp;=20 <BR>&gt; applied to<BR>&gt; &gt; psychology, as popularized&nbsp; by Attne= ave=20 after Garner &amp; Hake, has&nbsp; <BR>&gt; nothing at<BR>&gt; &gt; all to= do=20 with transmission of anything.&nbsp; When applied to&nbsp; <BR>&gt;=20 traditional<BR>&gt; &gt; absolute judgment (identification) experiments, f= or=20 example, it is&nbsp; <BR>&gt; merely an<BR>&gt; &gt; alternative measure o= f=20 short-term sensory memory.&nbsp; Which has&nbsp; <BR>&gt; nothing to=20 do<BR>&gt; &gt; with Claude Shannon's "general communications system".&nbs= p;=20 In your&nbsp; <BR>&gt; case, memory<BR>&gt; &gt; capacity is probably what= the=20 analysis will indicate.&nbsp; I am&nbsp; <BR>&gt; surprised that<BR>&gt; &= gt;=20 anyone uses the Garner-Hake information approach anymore;&nbsp; <BR>&gt;=20 mathematical<BR>&gt; &gt; psychologists (Duncan Luce, Donald Laming, Sandy= =20 MacRae, etc.)&nbsp; <BR>&gt; recognized<BR>&gt; &gt; its severe limitation= s=20 years ago and abandoned it.&nbsp; So&nbsp; <BR>&gt; (eventually) did<BR>&g= t;=20 &gt; acoustics users like Neff and Lutfi.&nbsp; The human being is NOT=20 a<BR>&gt; &gt; Shannon communication channel.&nbsp; Granted, the measures=20= that=20 result&nbsp; <BR>&gt; from<BR>&gt; &gt; processing the confusion matrix ar= e=20 covariance measures of a&nbsp; <BR>&gt; sort.&nbsp; In that<BR>&gt; &gt; c= ase,=20 ordinary covariance measures may provide a more meaningful&nbsp; <BR>&gt;=20= way=20 of<BR>&gt; &gt; analyzing your data.&nbsp; In response to your question, t= hen,=20 none of&nbsp; <BR>&gt; TRANS,<BR>&gt; &gt; TRANS/INPUT or TRANS/TI are=20 appropriate measures for your purposes.<BR>&gt; &gt; Information theory=20 doesn't work that way.&nbsp; Attneave got the math&nbsp; <BR>&gt;=20 right,<BR>&gt; &gt; but that had already been done; but Attneave's=20 interpretations&nbsp; <BR>&gt; (after Garner<BR>&gt; &gt; &amp; Hake) were= all=20 dead wrong.&nbsp; - Lance Nizami BSc (Physics) MSc&nbsp; <BR>&gt;=20 (Biomedical<BR>&gt; &gt; Engineering) PhD (Psychophysics), Decatur, GA=20 30030<BR>&gt; &gt;<BR>&gt; &gt; In a message dated 2/24/2009 4:54:12 A.M.=20 Eastern Standard Time,<BR>&gt; &gt; caro_jacquier@xxxxxxxx writes:<BR>&gt;= =20 &gt;<BR>&gt; &gt; Dear members,<BR>&gt; &gt;<BR>&gt; &gt; I would like to=20 analyse phonetic feature transmission (especially&nbsp; <BR>&gt;=20 voicing<BR>&gt; &gt; and place of articulation of french plosive consonant= s)=20 on the&nbsp; <BR>&gt; basis of<BR>&gt; &gt; individual confusion matrices.= My=20 study is about compressed speech<BR>&gt; &gt; perception in dyslexic=20 adults.<BR>&gt; &gt; I have already downloaded the software for informatio= n=20 transfer&nbsp; <BR>&gt; and SINFA<BR>&gt; &gt; analysis (Stuart Rosen ) an= d I=20 have run the analyses (on 32&nbsp; <BR>&gt; subjects).<BR>&gt; &gt; My fir= st=20 question is: which value is the more relevant in the&nbsp; <BR>&gt; report= =20 (TRANS,<BR>&gt; &gt; TRANS/INPUT or TRANS/TI) ?<BR>&gt; &gt; My second=20 question is about individual confusion matrices: Do I&nbsp; <BR>&gt; have=20= to=20 do<BR>&gt; &gt; one analyse per subject ?<BR>&gt; &gt;<BR>&gt; &gt; I hope= =20 that someone could help me.<BR>&gt; &gt; Thank you very much.<BR>&gt; &gt;= =20 Best regards,<BR>&gt; &gt;<BR>&gt; &gt; Caroline<BR>&gt; &gt;<BR>&gt;=20 &gt;<BR>&gt; &gt;<BR>&gt; &gt; ________________________________<BR>&gt; &g= t; A=20 Good Credit Score is 700 or Above. See yours in just 2 easy=20 steps!<BR>&gt;<BR>&gt; A Good Credit Score is 700 or Above. See yours in j= ust=20 2 easy steps!<BR><BR></FONT></BLOCKQUOTE></DIV></FONT><br/><font style=3D"= color:black;font:normal 10pt arial,san-serif;"> <hr style=3D"margin-top:10px= "/><b>A Good Credit Score is 700 or Above. <a href=3D"http://pr.atwola.com/p= romoclk/100126575x1218822736x1201267884/aol?redir=3Dhttp:%2F%2Fwww.freecredi= treport.com%2Fpm%2Fdefault.aspx%3Fsc%3D668072%26hmpgID%3D62%26bcd%3Dfebemail= footerNO62"> See yours in just 2 easy steps!</a></b></font></DIV></BODY></HT= ML> -------------------------------1235771543--


This message came from the mail archive
http://www.auditory.org/postings/2009/
maintained by:
DAn Ellis <dpwe@ee.columbia.edu>
Electrical Engineering Dept., Columbia University