{"title": "Development of Orientation and Ocular Dominance Columns in Infant Macaques", "book": "Advances in Neural Information Processing Systems", "page_first": 543, "page_last": 550, "abstract": null, "full_text": "Development of Orientation and  Ocular \nDominance Columns in Infant  Macaques \n\nKlaus  Obermayer \n\nHoward Hughes  Medical  Institute \n\nSalk-Institute \n\nLa Jolla, CA 92037 \n\nLynne Kiorpes \n\nCenter for  Neural Science \n\nNew  York  University \nNew  York,  NY  10003 \n\nGary  G.  Blasdel \n\nDepartment of Neurobiology \n\nHarvard  Medical  School \n\nBoston,  MA  02115 \n\nAbstract \n\nMaps of orientation preference and ocular dominance were recorded \noptically from the cortices of 5 infant macaque monkeys, ranging in \nage from 3.5 to 14 weeks.  In agreement with previous observations, \nwe  found  that basic features  of orientation and ocular  dominance \nmaps, as well  as correlations between them, are present  and robust \nby  3.5  weeks  of age.  We  did  observe  changes  in  the  strength  of \nocular dominance signals, as  well  as  in the spacing of ocular dom(cid:173)\ninance bands, both of which  increased steadily between  3.5 and  14 \nweeks  of age.  The  latter  finding  suggests  that  the  adult  spacing \nof ocular dominance bands depends on cortical growth in  neonatal \nanimals.  Since  we  found  no  corresponding  increase  in  the spacing \nof orientation  preferences,  however,  there  is  a  possibility  that  the \norientation preferences  of some cells  change as  the cortical surface \nexpands.  Since  correlations  between  the  patterns  of orientation \nselectivity  and ocular dominance  are  present  at  an  age,  when  the \nvisual system is still immature, it seems more likely that  their  de(cid:173)\nvelopment may be an innate process and may not require extensive \nvisual experience. \n\n543 \n\n\f544 \n\nObennayer, Kiorpes, and Blasdel \n\n1 \n\nINTRODUCTION \n\nOver  the  past  years,  high-resolution  images of the  simultaneous  representation  of \norientation selectivity  and  ocular  dominance  have  been  obtained  in  large  areas  of \nmacaque striate  cortex  using  optical  techniques  [3,  4,  5,  6,  12,  18].  These  studies \nconfirmed that ocular dominance and orientation  preference  are  organized  in  large \nparts in slabs.  While optical recordings of ocular dominance are in  accordance with \nprevious  findings,  it  turned  out  that  iso-orientation  slabs  are  much  shorter  than \nexpected,  and  that the orientation  map  contains  several  other  important elements \nof organization - singularities, fractures,  and saddle-points. \n\nA comparison between maps of orientation preference  and ocular dominance, which \nwere  derived  from  the  same region  of adult  monkey  striate  cortex,  showed  a  pro(cid:173)\nnounced  relationship  between  both  patterns  [5,  12,  13,  15,  17].  Fourier  analyses, \nfor  example, reveal  that orientation preferences  repeat  at closer  intervals along the \nocular dominance slabs than they do across  them.  Singularities were found to align \nwith  the  centers  of ocular dominance bands,  and  the  iso-orientation bands,  which \nconnect  them, intersect  the borders of ocular dominance bands preferably at angles \nclose  to 90\u00b0. \n\nGiven the fact  that these  relationships between  the  maps of orientation and ocular \ndominance are present in all maps recorded from adult macaques, one naturally won(cid:173)\nders  how  this organization matures.  If the ocular  dominance slabs were  to emerge \ninitially, for  example,  the  narrower  slabs of iso-orientation  might  later  develop  in \nbetween.  This might seem likely given the anatomical segregation which is apparent \nfor ocular dominance but not for orientation [9].  However,  this possibility is contra(cid:173)\ndicted by physiological studies that show normal, adult-like sequences of orientation \npreference  in  the  early  postnatal  weeks  in  macaque  when  ocular  dominance slabs \nare  still  immature [19].  The  latter findings  suggest  a  different  developmental  hy(cid:173)\npothesis;  that the organization into regions of different  orientation preferences  may \nprecede  or even  guide  ocular dominance formation.  A  third  possibility,  consistent \nwith  both  previous  results,  is  that  orientation  and  ocular  dominance  maps  form \nindependently  and align in later stages of development. \n\nIn  order  to  provide  evidence  for  one  or  the  other  hypothesis,  we  investigated  the \nrelationship  between  ocular  dominance  and  orientation  preference  in  very  young \nmacaque monkeys.  Results are  presented  in  the  remainder of this  paper.  Section 2 \ncontains  an  overview  about  the experimental  data,  and  section  3  relates  the  data \nto previous modelling efforts. \n\n2  ORIENTATION AND  OCULAR DOMINANCE \n\nCOLUMNS IN INFANT MACAQUES \n\n2.1  THE OVERALL STRUCTURE \n\nFigure  1 shows  the  map of orientation preference  (Fig.  1a)  and  ocular  dominance \n(Fig. 1 b) recorded from area 17 of a 3.5 week  old macaque. 1  Both maps look similar \n\n1 For all animals orientation  and ocular dominance  were  recorded  from  a  region close  to \n\nthe border  to area 18  and  close  to midline. \n\n\fDevelopment of Orientation and Ocular Dominance Columns in Infant Macaques \n\n545 \n\na \n\nb \n\nC \nFigure  1:  Spatial pattern  of orientation  preference  and ocular dominance recorded \nfrom area  17 of a  macaque, 3.5 weeks of age.  Figures  (a)  and  (b) show orientation \npreferences and ocular dominance bands within the same 3.1 mm x  4.3 mm large re(cid:173)\ngion of striate cortex.  Brightness values in  Fig.  (a)  indicate orientation preferences, \nwhere  the  interval  of  180\u00b0  is  represented  by  the  progression  in  colors  from  black \nto  white.  Brightness  values  in  Fig.  (b)  indicate  ocular  dominance,  where  bright \nand dark  denote  ipsi- and  contralateral  eye-preference.  respectively.  The data was \nrecorded  from  a  region  close  to the  border  to area  18  and  close  to midline.  Figure \n(c) shows an enlarged section of this map in  the preference  (left) and the  in  contour \nplot  (right)  representations.  Iso-orientation  lines  on  the  right  indicate  intervals of \n11.25\u00b0.  Letters  indicate  linear  zones  (L),  saddle  points  (H),  singularities  (S),  and \nfractures  (F). \n\nto maps which  have been  recorded  from  adults.  The orientation map exhibits all of \nthe  local elements  which  have  been  described  [12,  13]:  linear  zones,  saddle  points, \n\n\f546 \n\nObermayer, Kiorpes, and Blasdel \n\na \n\n::: \n\n')..r  = 741pm \n')..7;  = 612pm \n\n\u2022 \n00 \nA. = 724f.1m \n\n.... \n\nc \n\n1.0  - - - - - - - - - - - ,  \n\nb --0 \n\n41) \n.~ \n(ij \nE .... \n0 c: \n.... \n41) \n~ \n0 a. \n\n1.2 \n\n1.0 \n\n0.8 \n\n0.6 \n\n0.4 \n\n0.2 \n\n0.0 \n\nc: \no \n-.::::  ...-. \n0-0 \nc:  41)  0.5 \n.a.~ \ncCU \n.2 E \n100 \n... ... o o \n-\n41)(cid:173)\n\nc:  0.0 \n\n-0.5 \n\no \n\n200 \n\n400 \n\n600 \n\n800 \n\ndistance [~m] \n\n0 \n\n1 \n\n2 \n\n3 \n\n4 \n\n5 \n\nspatial frequency [l/mm] \n\nFigure  2:  Fourieranalysis of the  ori(cid:173)\nentation  map  shown  in  Figure  la. \n(a)  Complex  2D-Fouriertransform. \nEach pixel corresponds to one  Fouri(cid:173)\nermode  and  its  blackness  indicates \nthe  corresponding  energy.  A  dis(cid:173)\ntance  of  one  pixel  corresponds  to \nO.23/mm.  (b)  Power  as  a  function \nof radial  spatial  frequency.  (c)  Au(cid:173)\ntocorrelations  Sij  as  a  function  of \ndistance.  The  indices  i, j  E  {3,4} \ndenote the two cartesian coordinates \nof the  orientation  preference  vector. \nFor  details  on  the  calculation  see \n[13,  15]. \n\nsingularities, and fractures  (Fig.  lc).  The ocular  dominance map shows  its  typical \npattern of alternating bands. \n\nFigure  2a shows  the  result  of a  complex  2D  Fourier  transform  of the  orientation \nmap  shown  in  Figure  la.  Like  for  maps  recorded  from  adult  monkeys  [13]  the \nspectrum  is  characterized  by  a  slightly elliptical band of modes  which  is  centered \nat  the  origin.  The  major  axis  approximately  aligns  with  the  axis  parallel  to  the \nborder  to  area  18  as  well  as  with  the  ocular  dominance  bands.  Therefore,  like  in \nthe adults, the orientatiQn map is stretched perpendicular  to the ocular dominance \nbands, apparently to adjust to the wider spacing. \n\nWhen one  neglects  the slight  anisotropy of the  Fourier spectra one  can estimate a \npower  spectrum  by  averaging  the  squared  Fourier  amplitudes for  similar frequen(cid:173)\ncies.  The result  is  a  pronounced  peak  whose  location is  given  by  the  characteristic \nfrequency  of the orientation map (Fig. 2b).  As a  consequence,  autocorrelation func(cid:173)\ntions have  a  Mexican-hat shape (Fig. 2c), much like it has been  reported for  adults \n[13,  15]. \n\nIn summary, the basic features of the patterns of orientation and ocular dominance \nare  established  as  early  as  3.5  weeks  of age.  Data which  were  recorded  from  four \n\n\fDevelopment of Orientation and Ocular Dominance Columns in Infant Macaques \n\n547 \n\nTable 1:  Characteristic wavelengths (AOD) and signal strengths \u00abTOD) for  the ocular \ndominance pattern,  as  well  as  characteristic  wavelengths  p.op), density  of +180\u00b0-\nsingularities (p+),  density of -180\u00b0-singularities (p_),  total density  of singularities \n(p),  and percentage of area covered  by linear zones  (alin)  for  the orientation pattern \nas  a  function of age. \n\nage \n\n(weeks) \n\nUOD \n\n>'OD \n(}jm) \n\n>'OP \n(}jm) \n\np+ \n\n(mm- 2) \n\np-\n\n(mm- 2) \n\nP \n\n(mm- 2) \n\nalin \n\n(%  area) \n\n3.5 \n5.5 \n7.5 \n14 \n\nadult \n\n0.92 \n0.96 \n0.66 \n1.23 \n1.36 \n\n686 \n730 \n870 \n917 \n950 \n\n660 \n714 \n615 \n700 \n768 \n\n3.9 \n3.7 \n4.5 \n3.9 \n3.9 \n\n3.9 \n3.7 \n4.5 \n3.8 \n3.8 \n\n7.8 \n7.4 \n9.0 \n7.7 \n7.7 \n\n47 \n49 \n45 \n36 \n43 \n\nother infants ranging from  5.5  to 14  weeks  (not shown)  confirm the above findings. \n\n2.2  CHARACTERISTIC  WAVELENGTHS  AND  SIGNAL \n\nSTRENGTH \n\nA  more  detailled  analysis  of the  recorded  patterns,  however,  reveals  changes  of \ncertain  features  with  age.  Table  1 shows  the  changes  in  the  typical  wavelength  of \nthe  orientation and  ocular  dominance  patterns  as  well  as  the  (normalized)  ocular \ndominance signal strength  with  age.  The strength  of the  ocular  dominance signal \nincreases  by  a  factor  of 1.5  between  3.5  weeks  and  adulthood,  a  fact,  which  could \nbe explained by  the still ongoing segregation of fibers  within  layer  IV c. \n\nThe spacing of the ocular dominance columns increases  by approximately 30%  be(cid:173)\ntween  3.5  weeks  and  adulthood.  This change  in spacing  would  be  consistent  with \nthe growth of cortical surface  area during this period  [16]  if one assumes  that cor(cid:173)\ntex  grows  anisotropically  in  the  direction  perpendicular  to  the  ocular  dominance \nbands.  Interestingly,  the  characteristic  wavelengths of the  orientation  patterns  do \nnot  exhibit  such  an increase.  The  wavelengths  for  the  patterns  recorded  from  the \ndifferent  infants are close  to the  \"adult\" values.  More evidence for  a stable orienta(cid:173)\ntion pattern is provided by the fact, that the density of'singularities is approximately \nconstant  with age2  and that the percentage of cortical area covered  by linear zones \ndoes  neither  increase  nor  decrease.  Hence  we  are  left  with the puzzle  that at  least \nthe  pattern of orientation does  not follow  cortical  growth. \n\n2.3  CORRELATIONS  BETWEEN  THE  ORIENTATION  AND \n\nOCULAR DOMINANCE  MAPS \n\nFigure 3 shows a contour plot representation of the pattern of orientation preference \nin  overlay  with  the  borders  of  the  ocular  dominance  bands  for  the  3.5  week  old \nanimal.  Iso-orientation  contours  (thin  lines)  indicate  intervals  of  15\u00b0.  Thick  lines \nindicate  the  border  of the  ocular  dominance  bands.  From  visual  inspection  it  is \n\n2Note  that  both  types of singularities  appear  in  equal  numbers. \n\n\fS48 \n\nObennayer. Kiorpes. and Blasdel \n\nFigure  3:  Contour  plot  representa(cid:173)\ntion of t.he  orient.a.t.ion  map shown  in \nFigure  la in overlay with the borders \nof the ocular dominance bands taken \nfrom  Figure  1 b.  Iso-orient.ation  lines \n(thin  lines)  indicate  intervals of 15\u00b0. \nThe borders of the ocular dominance \nbands are indicated  by  thick  lines. \n\nalready  apparent  that singularities have a  strong tendency  to  align with the center \nof the  ocular  dominance  bands  (arrow  1)  and  that  in  the  linear  zones  (arrow  2), \nwhere  iso-orientation bands exist,  these  bands intersect  ocular dominance bands at \nangles close  to 90\u00b0  most of the time. \n\nTable  2  shows  a  quantitative  analysis  of the  local  intersection  angle.  Percentage \nof area  covered  by  linear  zones  (cf.  [12]  for  details  of the  calculation)  is  given  for \nregions,  where  orientation  bands  intersect  ocular  dominance  bands  within  18\u00b0  of \nperpendicular,  and  regions  where  they  intersect  within  18\u00b0  of parallel.  For  all  of \nthe  animals investigated  the  percentages  are  two  to  four  times  higher  for  regions, \nwhere  orientation  bands  intersect  ocular  dominance  bands  at  angles  close  to  90\u00b0, \nmuch  like  it  has  been  observed  in  adults  [12].  In  particular  there  is  no  consistent \ntrend  with  age: \nthe  correlations  between  the  orientation  and  ocular  dominance \nmaps are established  as early as 3.5 weeks  of age. \n\nage \n\n(weeks) \n\nperp \naUn \n(%area) \n\npar \na'in \n\n(%  area) \n\n3.5 \n5.5 \n7.5 \n14 \n\nadult \n\n15.9 \n12.2 \n13.3 \n12.4 \n18.0 \n\n4.1 \n6.8 \n6.2 \n3.7 \n2.7 \n\nTable 2:  Percentage of area covered  by \nlinear zones as a function of age for  re-\ngions,  where  orientation  bands  inter-\nsect  ocular  dominance  bands  within \n18\u00b0  of  perpendicular  (af/::'P) ,  and  re-\ngions  where  they  intersect  within  18\u00b0 \nof parallel (afi~) (cf.  [12]  for  details of \nthe calculation). \n\n\fDevelopment of Orientation and Ocular Dominance Columns in Infant Macaques \n\n549 \n\n3  CONCLUSIONS  AND  RELATION  TO  MODELLING \n\nIn  summary, our  results  provide  evidence  that  the  pattern  of orientation  is  estab(cid:173)\nlished at a  time when  the pattern  of ocular dominance is still developing.  However, \nthey  provide also evidence  for  the fact  that  the  pattern  of orientation is  not  linked \nto cortical growth.  This latter finding still  needs  to be firmly established  in  studies \nwhere  the development of orientation is followed  in one and the same animal.  But if \nit is taken seriously the consequence  would be  that orientation preferences  may shift. \nand that pairs of singularities are formed.  The early  presence  of strong correlations \nbetween  both  maps indicate  that  the  development  of orientation  and  ocular  dom(cid:173)\ninance are  not  independent  processes.  Both  patterns  have  to adjust.  to  each  other \nwhile  cortex  is  growing.  It,  therefore,  seems  as  if the  third hypothesis  is  true  (see \nIntroduction)  which states  that both patterns develop  independently  and  adjust to \neach other  in  the  late stages of development.  As  has  been shown  in  [13,  15]  and  is \nsuggested  in  [7,  14]  these  processes  are  certainly  in  the  realm  of models  based  on \nHebbian learning. \n\nMany features  of the orientation and ocular dominance maps are  present  at an age \nwhen  the visual  system of the  monkey  is  still  immature [8,  11].  In  particular,  they \nare  present  at  a  time  when  spatial  vision  is  strongly  impared.  Consequently,  it \nseems  unlikely that the  development of these features  as  well  as of the  correlations \nbetween  both  patterns  requires  high  acuity  form  vision,  and  models  which  try  to \npredict  the  structure  of these  maps  from  the  structure  of visual  images  [1,  2,  10] \nhave  to  take  this  fact  into  account.  The  early  development  of orientation  prefer(cid:173)\nence  and its correlations with ocular dominance make it also seem  more likely  that \ntheir  development  may me an  innate process  and  may not  require extensive  visua.l \nexperience.  Further experiments,  however,  are  needed  to settle  these  questions. \n\nAcknowledgements \n\nThis work was funded in part by the Klingenstein Foundation, the McKnight Foun(cid:173)\ndation,  the  New  England  Primate  Research  Center  (P51RR0168-31),  the  Seaver \nInstitute,  and  the  Howard  Hughes  Medical  Institute.  We  thank  Terry  Sejnowski, \nPeter  Dayan,  and  Rich  Zemel  for  useful  comments on  the  manuscript.  Linda  As(cid:173)\ncomb, Jaqueline  Mack,  and Gina Quinn provided excellent  technical  assistance. \n\nReferences \n\n[1]  H.  G.  Barrow  and  A.  J.  Bray.  Activity  induced  color  blob  formation .  In \nI.  Alexander  and  J.  Taylor, editors,  Artificial  Neural  Networks  II,  pages  5-9. \nElsevier  Publishers,  1992. \n\n[2]  H.  G. Barrow and A.  J. Bray.  A model of the adaptive development of complex \ncortical cells.  In I. Alexander and J. Taylor, editors,  A rtificial Neural Networks \nII,  pages  1-4. Elsevier  Publishers,  1992. \n\n[3]  E.  Bartfeld and A.  Grinvald.  Relationships between orientation-preference pin(cid:173)\n\nwheels,  cytochrome  oxidase  blobs,  and  ocular-dominance columns  in  primate \nstriate cortex.  Proc.  Nail.  Acad.  Sci.  USA,  89:11905-11909, 1992. \n\n\f550 \n\nObennayer, Kiorpes, and Blasdel \n\n[4]  G.  G.  Blasdel.  Differential imaging of ocular dominance and orientat.ion selec(cid:173)\n\ntivity in  monkey striate cortex.  J.  Neurosci.,  12:3117-31:>8,  1992. \n\n[5]  G.  G.  Blasdel.  Orientation  selectivity,  preference,  and  continuity  in  monkey \n\nstriate cortex.  J.  Neuroscl.,  12:3139-3161, 1992. \n\n[6]  G.  G.  Blasdel and G.  Salama.  Voltage sensitive dyes  reveal  a  modular organi(cid:173)\n\nzation  in  monkey striate cortex.  Nature,  321:579-585, 1986. \n\n[7]  R.  Durbin  and  G.  Mitchison.  A  dimension  reduction  framework  for  under(cid:173)\n\nstanding cortical  maps.  Nature,  343:644-647,  1990. \n\n[8]  L.  Kiorpes  and  T.  Movshon.  Behavioural  analysis  of visual  development.  In \nJ.  R.  Coleman,  editor,  Development  of Sensory  Systems  m  Mammals,  pages \n125-154. John  Wiley,  1990. \n\n[9]  S. LeVay, D.  H.  Hubel, and T. N. Wiesel. The development of ocular dominance \ncolumns in normal and visually deprived monkeys.  J.  Compo  Neurol.,  191:1-51, \n1980. \n\n[10]  Y.  Liu  and  H.  Shouval.  Principal  component  analysis  of natural  images - an \n\nanalytic solution.  Preprint. \n\n[11]  T.  Movshon  and  L.  Kiorpes.  Biological  limits  on  visual  development.  in  pri(cid:173)\n\nmates.  In  K.  Simon,  editor,  Handbook  of Infant  Vision.  Oxford  University \nPress,  1993.  in  press. \n\n[12]  K.  Obermayer  and  G.  G .  Blasdel.  Geometry of orientation  and  ocular domi(cid:173)\n\nnance  columns in  monkey striate cortex.  J.  Neurosci.,  13:4114-4129, 1993. \n\n[13]  K.  Obermayer, G. G. Blasdel, and K.  Schulten.  A statistical mechanical a.naly(cid:173)\n\nsis of self-organization and pattern formation during the development of visual \nmaps.  Phys.  Rev.  A15,  45:7568-7589,  1992. \n\n[14]  K.  Obermayer,  H.  Ritter,  and  K.  Schulten.  A  principle  for  the  forma.t.ion  of \nthe  spatial  structure  of cortical  feature  maps.  Proc.  Natl.  Acad.  Sci.  USA, \n87:8345-8349, 1990. \n\n[15]  K.  Obermayer,  K.  Schulten,  and  G.  G.  Blasdel.  A  comparison  of  a  neural \nnetwork  model  for  the  formation  of brain  maps  with  experimental  data.  In \nD.  S.  Touretzky  and  R.  Lippman,  editors,  Advances  in  Neural  Information \nProcessing Systems  ./,  pages 83-90. Morgan  Kaufmann  Publishers,  1992. \n\n[16]  D.  Purves  and  A.  LaMantia.  Development  of  blobs  in  the  visual  cortex  of \n\nmacaques.  J.  Compo  Neurol.,  332:1-7,1993. \n\n[17]  N.  Swindale.  A  model  for  the  coordinated  development  of columnar systems \n\nin  primate striate cortex.  Bioi.  Cybern.,  66:217-230, 1992. \n\n[18]  D. Y. Tso, R.  D. Frostig, E.  E.  Lieke,  and A.  Grinvald.  Functional organization \nof primate visual  cortex  revealed  by  high  resolution optical  ima.ging.  Science, \n249:417-420,  1990. \n\n[19]  T.  N.  Wiesel  and  D.  H.  Rubel.  Ordered  arrangement  of orientation  columns \n\nin  monkeys lacking  visual experience.  J.  Compo  Neurol.,  158:307-318, 1974. \n\n\f", "award": [], "sourceid": 868, "authors": [{"given_name": "Klaus", "family_name": "Obermayer", "institution": null}, {"given_name": "Lynne", "family_name": "Kiorpes", "institution": null}, {"given_name": "Gary", "family_name": "Blasdel", "institution": null}]}