{"title": "An Information-Theoretic Approach to Deciphering the Hippocampal Code", "book": "Advances in Neural Information Processing Systems", "page_first": 1030, "page_last": 1037, "abstract": null, "full_text": "An Information-Theoretic  Approach to \n\nDeciphering the Hippocampal Code \n\nWilliam E.  Skaggs  Bruce L.  McNaughton  Katalin M.  Gothard \n\nCenter for  Neural Systems,  Memory,  and Aging \n\nEtan J.  Markus \n\n344 Life  Sciences  North \nUniversity of Arizona \n\nTucson AZ  85724 \n\nbill@nsma.arizona.edu \n\nAbstract \n\nInformation  theory  is  used  to  derive  a  simple  formula  for  the \namount of information conveyed by the firing rate of a neuron about \nany  experimentally measured variable or  combination of variables \n(e.g.  running speed,  head  direction,  location of the  animal, etc.). \nThe  derivation  treats  the  cell  as  a  communication channel  whose \ninput is the measured variable and whose output is the cell's spike \ntrain.  Applying the formula,  we  find  systematic differences  in  the \ninformation  content  of hippocampal  \"place  cells\"  in  different  ex(cid:173)\nperimental conditions. \n\n1 \n\nINTRODUCTION \n\nAlmost any neuron will respond to some manipulation or other by changing its firing \nrate,  and this change in firing can convey information to downstream neurons.  The \naim of this article is to introduce a very simple formula for the average rate at which \na  cell  conveys  information in  this  way,  and  to show  how  the formula is  helpful  in \nthe  study  of the  firing  properties  of cells  in the  rat  hippocampus.  This  is  by  no \nmeans the first  application of information theory to the study of neural coding; see \nespecially Richmond and Optican (1990).  The thing that particularly distinguishes \n\n1030 \n\n\fAn  Information-Theoretic Approach  to Deciphering the  Hippocampal  Code \n\n1031 \n\nour approach  is  its simplemindedness. \nTo get the basic idea, imagine we are recording the activity of a neuron in the brain \nof a rat, while the rat is wandering around randomly on a circular platform. Suppose \nwe  observe  that the  cell  fires  only when  the rat is  on  the left  half of the platform, \nand that it fires  at a constant rate everywhere on the left half; and suppose that on \nthe whole the rat spends half of its time on the left half of the platform.  In this case, \nif we  are prevented  from seeing  where  the rat is,  but are informed that the neuron \nhas just this  very  moment fired  a  spike,  we  obtain thereby  one  bit of information \nabout  the  current  location of the  rat.  Suppose  we  have  a  second  cell,  which  fires \nonly in the southwest  quarter of the platform; in this case a spike would give us two \nbits of information.  If there  were  in addition a small amount of background firing, \nthe information would  be slightly less  than two bits.  And so  on. \nGoing back to the cell that fires  everywhere on the left half of the platform, suppose \nthat when it is active, it fires at a mean rate of 10 spikes per second.  Since it is active \nhalf the time, it fires  at  an overall mean rate of 5 spikes  per second.  Since  a  spike \nconveys one bit of information about the rat's location, the cell's spike train conveys \ninformation at an average rate of 5 bits per second.  This does not mean that if the \ncell  is  observed for  one second,  on average  5 bits will be obtained-rather it means \nthat if the cell  is  observed for  a sufficiently short time interval dt, on average  5dt \nbits  will  be  obtained.  In  20  milliseconds,  for  example,  the  expected  information \nconveyed by the cell  about the location of the rat will be very nearly 0.1  bits.  The \nlonger the time interval over which the cell is observed,  the more redundancy in the \nspike train,  and hence  the farther below  5dt the total information falls. \nThe formula that leads to these  numbers is \n\n1= l.\\(X)  log2 .\\~)  p(x)dx, \n\n(1) \n\nwhere  I  is  the information rate  of the  cell in bits per second,  x  is  spatial location, \np( x) is  the probability density for the rat being at location x, .\\( x)  is the mean firing \nrate when the rat is  at location x,  and .\\ = Jz .\\(x)p(x)dx is the overall mean firing \nrate of the cell.  The derivation of this formula appears in the final section.  (To our \nknowledge  the formula, though very simple, has not previously been published.) \n\nNote that, as  far  as the formula is  concerned,  there is nothing special about spatial \nlocation:  the  formula  can  equally  well  be  used  to  define  the  rate  at  which  a  cell \nconveys  information  about  any  aspect  of the  rat's  state,  or  any  combination  of \naspects.  The only mathematical requirement 1  is that the rat's state x  and the spike \ntrain of the cell both be stationary random variables, so that the probability density \np( x)  and the expected firing  rate .\\( x)  are  well-defined. \nThe information rate given  by formula (1)  is  measured  in  bits  per  second.  If it is \ndivided by the overall mean firing rate.\\ of the cell  (expressed in spikes per second), \nthen a different kind of information rate is obtained, in units of bits per spike-let us \ncall it the information per spike.  This is  a measure of the specificity of the cell:  the \nmore  \"grandmotherish\"  the cell,  the more information per spike.  For a  population \n\nlOther than obvious  requirements of integrability  that are sure to be fulfilled  in natural \n\nsituations. \n\n\f1032 \n\nSkaggs,  McNaughton, Gothard,  and Markus \n\nof cells,  then,  a  highly distributed  representation  equates  to little information per \nspike. \n\n. \n' .. :. .. ,'. \n\n.' \n.  .  . \n. I, \n\u2022 \n\n' \u2022 \n\n\u2022 \n\n\u2022 \n\n.. \u2022\u2022\u2022 \n\n\u2022  .: \u2022\u2022\u2022\u2022 \n\n\u2022\u2022\u2022  \u2022 \n\n\u2022 \n\n.11 \n\n.. ':,'.  .  J  ..  ' ': \n'.  .. \n.  ..'.. \n\nI  \", 'I: .'  ,I \n\nI \n\n\"I \u2022\u2022  II'~  ,'I'  . .  II \n...'  IJ \n\u2022 \n\nI  I.'  \u2022 .  \n\n, \"   . \" .  \n\n'1 \n\n\u2022 \n\n. '  \n\n\u2022 \u2022 \u2022 \u2022 \u2022  1 \n\n___ \n\nI I   . , '   \u2022 \u2022 \u2022 \u2022  \n\n'  \u2022\u2022 - . .  \n\nI  '. \n\nI \n\n. '   I  \u2022\n\u2022 \u2022  \n\n\u2022  I .  \u2022 \n/ \"  \n\n\u2022 \nI \n\n' \n\n\u2022 \u2022  I I  \n\n\u2022 \n\n. . . .  \n\n\u2022  \u2022 \u2022   I I .  \n\n. '  \n\nI '   .1  '.. \n. \n\n\u2022  \u2022 \n\u2022.... .1  .':  .  ,.'. \n\n:..  \u2022\u2022\u2022 \n' . .  \u2022 \n\u2022 \n\n\u2022 \n\n\u2022 \n\n\u2022 \n\n.. \n\n.', \n\n\u2022\u2022 ::.' \n\n.. ~.'  \"_ \nII:. \nL_ \nI I : ' .   I I   ~'.' \" ' :  \nII. I .   \u2022  ,',  .' \n\n\u2022 \n\n..  ILl. \n\u2022  ' .   \u2022 \u2022 \u2022 \u2022 \u2022 \u2022  I  \u2022 \n' \n\n\u2022 \u2022  '  \u2022 \u2022   \u2022  I. \n.1  ' . '   .  I.' \n\n\" \n\nI \n\nII  \u2022 \n\n. .  I \n\n\u2022 \n\n... \n\u2022 \u2022 \u2022  \n\u2022 \n\u2022 \n\n\u2022 \n\n\u2022 \n\n\u2022 \n\n\u2022 \n\n: \n\n\u2022 \n\nI \n\n, \"  \n\n\u2022 \n\n? \n\nI \n\n' .  \n\n\u2022 \u2022 \u2022  \n\n\u2022 \n\n\u2022  ' .   \u2022 \n\n\u2022\u2022  ' , . '  \n\n'. \n\n\"1   I . '  I .  \n\n\u2022  ' . '   . . . .   \" . '   . . . . . . : . .  \u2022 \n\n. . .1 .  \nI  . \"   \u2022\u2022\u2022\u2022 \nI \n\n.1 \u2022\u2022\u2022 I .. I I   \u2022 \n\" ,1 '   . , '1 , '   , ' . '   \u2022 \u2022 \u2022 \u2022 \u2022 \u2022 \u2022 \u2022 \u2022 \u2022   I .  \n\u2022 \u2022   I' \n\u00b7  .:'  ....  \",  I\u00b7\u00b7\u00b7\u00b7\u00b7  \u00b7' \u2022. 1.\u00b7\u00b7 ....\u2022.... \n...  \u2022\u2022\u2022 \n\u2022  I \nI \n\u2022  I '  \n'.' ':\".  . \n\u00b7 II ..  .....  .' . \n. I...: \nI.  '.. .\u00b7 .... 1._.\u00b7..  . .'  ..... .. '. \\  '. \n\u2022  '.' .1, .~:. '. I.\". \n\u2022 \nI  \":.'  I. I .. 1  \u2022\u2022\u2022\u2022 I' .'  '.  '. \u2022\u2022  I  :.:  I... I '.  ..,... \nI \u00b7   I ' .   \u2022\u2022\u2022\u2022\u2022\u2022 \n.  ..... \nI  J  I \n. . . . .  '--\nI  I.....  t .. 1  _ ' \u2022\u2022\u2022 :1.' \n\u2022 \n\u2022 \nII. \n~I \n'.  '.:\" \n\u2022 \n. I  I \nI \nCO' \nII.'. \n....  . \n\u2022 \n\u2022 ... \n\u2022  I. \nI..  .;  \"  I:. :~. . \nI \n'1  \u2022 \u2022  \n\n..' I..  I  I \n.'. \n..1 \n\n.  .. I...  . '. \n\nI \n. \u2022 \n.1.  \u2022 I  .' \n\u2022  '.l!)' \n\n'1  \u2022 \u2022 \u2022   I .  \nI .  \n\nI,.~.  .1' \u2022 \n\n\u2022  I'\" \nI \n\n\u2022  I  I.'  .. \n\n\u2022  II \n\n,\"  I.' \n\n\u2022 \u2022 \u2022 \u2022 \u2022  _ \n\n\u2022  '.  I\"  II.  I \n\n\u2022 \n\u2022  \u2022 \n\n',  \u2022 \u2022  ' \n\n\u2022  I.'  .'  \u2022 \n\n\u2022 \u2022 \u2022 \u2022 \u2022 \u2022 \u2022 \u2022  \n\n\u2022  , \\ . . .  \n\n\u2022  ........... \n\n'. \n\n\u2022 \u2022\u2022\u2022 \n\n\u2022  . . . . .  \n\nI \n\n\u2022 \n\n.\\ \n\n' .   ' .  \n\nI \n\n\u2022 \n\n1 .1  \n\n\u2022 \n\n\u2022 \n\n.\"  \u2022 \n\n.\" \n\nI \n\n\u2022 \n\n\u2022 \n\n\u2022 \n\nI \n\n\u2022 \n\n\u2022 \n\n\u2022 \n\n\u2022 \n\n\u2022 \n\nI I  \n\n-\n\n\u2022 \n\nI \n\n\u2022 \n\n. \n\n-..; :- \u2022  I'\u00b7~  ~~~~\\H~,,-.... ~g.!~lt\u00a31~ \n...  --;-\n\n~~~ \n\nFigure 1:  \"Spot plot\"  of the activity of a single pyramidal cell in the hippocampus \nof a  rat, recorded  while the rat foraged for food pellets inside a small cylinder.  The \ndots  show  locations  visited  by  the  rat,  and  the  circles  show  points  where  the  cell \nfired-large circles mean that several spikes occurred within a short time.  The lines \nindicate which direction the  rat was facing when the cell fired.  The plot represents \n29  minutes of data, during which  the cell fired  at an overall mean rate of 1.319 Hz. \n\nConsider,  as  an  example,  a  typical  \"place  cell\"  (actually  an  especially  nice  place \ncell) from the CAl layer of the hippocampus of a rat-Figure I  shows a  \"spot plot\" \nof the activity of the cell as the rat moves around inside a 76  cm diameter cylinder \nwith high, opaque walls, foraging for  randomly scattered food pellets.  This cell,  like \nmost pyramidal cells in CAl, fires  at a relatively high rate (above  10  Hz)  when the \nrat is  in a specific small portion of the environment-the \"place field\"  of the cell(cid:173)\nbut  at  a  much  lower  rate  elsewhere.  Different  cells  have  place  fields  in  different \nlocations;  there  are  no  systematic  rules  for  their  arrangement,  except  that  there \nmay be a tendency for neighboring cells  to have nearby place fields.  The activity of \nplace  cells  is  known  to be  related  to more than just place:  in some  circumstances \nit  is  sensitive  to  the  direction  the  rat  is  facing,  and  it  can  also  be  modulated by \nrunning speed,  alertness,  or other aspects  of behavioral state.  The dependence  on \n\n\fAn  Information-Theoretic Approach  to Deciphering the  Hippocampal  Code \n\n1033 \n\nhead  direction  has  given  rise  to  a  certain  amount of controversy,  because  in  some \ntypes  of environment it is  very strong,  while in others it is virtually absent. \nTable  1 gives  statistics for  the  amount of information conveyed  by  this cell  about \nspatial location, head direction, running speed,  and combinations of these variables. \nNote  that  the  information  conveyed  about  spatial  location  and  head  direction  is \nhardly  more  than the information conveyed  about spatial location  alone-the dif(cid:173)\nference  is  well  within  the  error  bounds  of the  calculation.  Thus  this  cell  has  no \ndetectable directionality.  This seems to be typical of cells  recorded  in unstructured \nenvironments. \n\nTable 1:  Information conveyed  by the cell whose  activity is  plotted in  Figure  1. \n\nVARIABLES \n\nLocation \n\nHead  Direction \nRunning Speed \n\nLocation  and Head  Direction \nLocation  and Running Speed \n\nINFO \n\n2.40  bits/sec \n0.48 bits/sec \n0.03 bits/sec \n2.53 bits/sec \n2.36 bits/sec \n\nINFO  PER SPIKE \n\n1.82 bits \n0.37  bits \n0.02  bits \n1.92 bits \n1.79 bits \n\nThe  information-rate measure  may be  helpful  in understanding  the  computations \nperformed  by  neural  populations.  Consider  an  example.  Cells  in  the  CA3  and \nCAl  regions  of the  rat  hippocampal formation  have  long  been  known  to  convey \ninformation about  a  rat's spatial location  (this  is  discussed  in  more detail below). \nData from our lab suggest that, in a given environment, an average CA3 cell conveys \nsomething in the neighborhood of 0.1 bits per second about the rat's position-some \ncells  convey  a good  deal more information than this,  but many are virtually silent. \nCells in CAl receive most of their input from cells in CA3; each gets on the order of \n10,000 such inputs.  Question:  How long must the integration time of a CAl cell be \nin  order for  it to form  a good estimate of the rat's location?  Answer:  With  10,000 \ninputs, each  conveying on average  0.1  bits per second,  the cell  receives  information \nat  a  rate of 1000  bits per second,  or  1 bit per millisecond, so  in 5-10 msec the  cell \nreceives  enough information to form a  moderately precise estimate of location. \n\n2  APPLICATIONS \n\nWe  now  very  briefly  describe  two  experimental studies  that have found  differences \nin  the spatial information content of rat hippocampal activity under  different  con(cid:173)\nditions.  The  methods  used  for  recording  the  cells  are  described  in  detail  in  Mc(cid:173)\nNaughton  et  al  (1989)-to summarize,  the  cells  were  recorded  with  stereotrodes, \nwhich  are  twisted  pairs  of electrodes,  separated  by  about  15  microns  at  the  tips, \nthat  pick  up  the  extracellular  electric  fields  generated  when  cells  fire.  A  single \nstereotrode  can detect  the activity of as  many as  six or seven distinct hippocampal \ncells;  spikes from different  cells  can be separated on the basis of their amplitudes on \nthe  two  electrodes,  as  well  as  other  differences  in  wave shape.  The location of the \nrat was tracked using arrays of LEDs attached to their heads and a video camera on \nthe ceiling.  Spatial firing rate maps for each cell were constructed using an adaptive \nbinning technique designed  to minimize error (Skaggs and McNaughton,  submitted), \n\n\f1034 \n\nSkaggs,  McNaughton,  Gothard,  and Markus \n\nand  information rates  were  calculated  using  these  firing  rate  maps.  As  a  control, \nthe spike train was  randomly time-shifted relative to the sequence of locations; this \nwas  done  100 times,  and the cell was deemed to have significant spatial dependence \nif its information rate was  more than 2.29  standard  deviations  above  the  mean of \nthe  100  control information rates. \n\n2.1  EXPERIMENT:  PROXIMAL VERSUS  DISTAL VISUAL  CUES \n\nIn this study (a preliminary account of which appears in Gothard  et  al (1992\u00bb, the \nactivity of place cells was recorded successively in two environments, the first a 76 em \ndiameter cylinder with four patterned cue-cards on the high, opaque gray wall, the \nsecond  a  cylinder  of the  same  shape,  but  with  a  low,  transparent  plexiglass  wall \nand four  patterned cue-cards on the distant black walls of the recording room.  The \ntwo environments thus had the same shape, and from  any given point were  visually \nquite  similar;  the  difference  is  that  in  one  all  of the  visual  cues  were  proximal to \nthe rat,  while in the other many of them were  distal. \n\nDISTAL CUES \n\nPROXIMAL CUES \n\n.:.,,:  :.:; \n\n. . . \n\n... \n\n. .. \n\n;:;: \n.. .; \n\n. .. \n\n. . \n\n\" \n\n.\n\n: : : \n\n. \n:.  : \n\nFigure  2:  Firing  rate  maps of four  simultaneously recorded  cells,  in the  distal  cue \nenvironment  (top)  and  proximal cue  environment  (bottom).  The scale  is  identical \nfor  all plots; black  ~ 5 Hz. \n\nFifty  cells  were  recorded  with  robust  place-dependent  firing  in  one  or  the  other \ncylinder.  There  was  no  discernable  relationship  between  place  fields  in  the  two \nenvironments-a cell  having  a  place field  in  the  proximal cue  environment  might \nbe  nearly silent in  the distal  cue environment,  and even if it did fire,  its place field \nwould  be  in  a  different  location.  (Figure  2 shows  firing  rate  maps for  four  of the \ncells.)  A substantially higher fraction of the cells had place fields in the proximal cue \nenvironment, and overall the average information per second was almost 50% higher \n\n\fAn  Information-Theoretic Approach  to  Deciphering the Hippocampal Code \n\n1035 \n\nin  the  proximal cue  environment.  For  the  cells  possessing  fields,  the  information \nper  spike  was  significantly higher  in  the  proximal cue  environment,  meaning that \nplace fields  were  more compact. \nThese  results  indicate  that  in  the  proximal  cue  environment,  spatial  location  is \nrepresented  by the hippocampal population more precisely,  and by a larger pool of \ncells,  than in  the  distal  cue  environment.  The  most likely  explanation  is  that,  at \nleast in the absence  of local cues,  the configuration of visual landmarks controls the \nactivity of the place cell  population. \n\n2.2  EXPERIMENT:  LIGHT VERSUS  DARK \n\nVisual  cues  have  a  great  deal  of influence  on  place  fields,  but  they  are  not  the \nonly important factor;  in fact,  some hippocampal cells maintain place fields  even in \ncomplete darkness  (McNaughton  et  a/.,  1989b; Quirk et  a/., 1990).  This experiment \n(Markus  et  a/., 1992)  was  designed  to examine how lack  of visual cues  changes  the \nproperties  of place  fields.  Rats  traversed  an  eight-arm  radial  maze  for  chocolate \nmilk reward,  with  the  room lights  being turned  on  and  oft' on  alternate  trials.  (A \ntrial consisted  of one visit  to each  of the  eight  arms of the maze.)  Figure 3 shows \nfiring  rate maps for four  simultaneously recorded  cells. \n\n.tt:\u00b7: \n'::::.::;.: \n\n:/::;. .. : .... : \n\nLIGHT \n\nDARK \n\nFigure 3:  Firing rate maps of four  simultaneously recorded  cells,  with room lights \nturned on (top)  and off (bottom).  The scale is  identical for  all plots; black  ~ 5 Hz. \n(The loops  at the ends of the arms are  caused  by the rat turning around there.) \n\nThe  most salient  effect  was  that  a  much larger fraction  of cells showed  significant \nspatially  selective  firing  in  the  light  than  in  the  dark:  35%  as  opposed  to  20%. \nHowever,  the  average  information  per  second  decreased  only  by  15%  in  the  dark \nas  compared  to the light, from  0.326  bits per second  in  the light to 0.278  bits per \n\n\f1036 \n\nSkaggs,  McNaughton,  Gothard,  and Markus \n\nsecond  in  the  dark.  (These  are  overestimates of the population  averages,  because \ncells silent in both light and dark were not included in the sample.) \nInterestingly, the drop in information content from light to dark seemed to be much \nsmaller than the drop from proximal cues  to distal cues  in the previous experiment. \nA  major difference  between  the  two  experime:nts  is  that,  in  the  eight-arm  maze, \ntactile cues  potentially give a  great  deal of information about spatial location,  but \nin  a  cylinder  they  serve  only  to  distinguish  the  center  from  the  wall.  While  it \nis  dangerous  to compare  the  two  experiments,  which  differed  methodologically in \nseveral  ways,  the results  suggest  that tactile cues  can have a  very  strong influence \non  hippocampal firing,  at least when  visual cues  are absent. \n\n3  THEORY \n\nThe information-rate formula (1)  is derived by considering a neuron  as  a  \"channel\" \n(in  the  information-theoretic sense)  whose  input  is  the  spatial location of the rat, \nand  whose  output  is  the  spike  train.  During  a  sufficiently short  time interval  the \nspike train is  effectively  a binary random variable (Le.  the only possibilities  are  to \nspike once or not  at all), and the probability of spiking is determined by the spatial \nlocation.  The event of spiking may be indicated by a random variable S whose value \nis  1 if the cell spikes and 0 otherwise.  If the environment is partitioned into a set of \nnonoverlapping bins,  then spatial location may be  represented  by an integer-valued \nrandom variable X  giving the index of the currently occupied  bin. \nIn  information theory,  the  information conveyed  by  a  discrete  random variable X \nabout  another  discrete  random variable Y, which  is  identical to the mutual infor(cid:173)\nmation of X  and Y,  is  given  by \n\nwhere  Z i  and Yi  are the  possible values of X  and Y,  and pO  is  probability. \nIf Aj  is  the mean firing  rate when the rat is in bin j, then the probability of a spike \nduring a brief time interval tl.t  is \n\nP(S= 11X =j) =  Ajtl.t. \n\nAlso,  the overall probability of a spike is \n\nP(S=1) = Atl.t, \n\nwhere \n\nwith Pi  =  P(X =j). \nAfter  these  expressions  are  plugged  in  to  the  equation  for  I(Y IX)  above,  it  is  a \nmatter of straightforward algebra,  using power series  expansions of logarithms and \nkeeping only lower order terms, to derive  a  discrete  approximation of equation  (1). \n\n\fAn  Information-Theoretic Approach  to  Deciphering the Hippocampal Code \n\n1037 \n\n4  DISCUSSION \n\nIn  many situations,  neurons  must  decide  whether  to fire  on  the  basis  of relatively \nbrief  samples  of  input,  often  100  milliseconds  or  less.  A  cell  cannot  get  much \ninformation from a single input in such  a short time, so  to achieve precision it needs \nto integrate many inputs.  Formula (1)  provides a measure of how much information \na  single input  conveys  about  a  given  variable in such  a  brief time interval. \n\nThe  formula  can  be  applied  to  any  type  of  cell  that  uses  firing  rate  to  convey \ninformation.  The  only  requirement  is  to  have  enough  data  to  get  good,  stable \nestimates  of firing  rates.  In  practice,  for  a  hippocampal cell  having  a  mean firing \nrate  of  around  0.5  Hz  in  an  environment,  twenty  minutes of data is  adequate  for \nmeasuring position-dependence;  and for  a  \"theta cell\"  (an  interneuron,  firing  at  a \nconsiderably  higher  rate),  very  clean measurements  are possible. \n\nWe  have  used  the  measure  in  the  study  of hippocampal  place  cells,  but  it  might \nactually  work  better  for  some  other  types.  The  problem  with  place  cells  is  that \nthey  fire  at  low  overall  rates,  so  it  is  time-consuming to  get  an  adequate  sample. \nCortical pyramidal cells  often  have mean rates  at least  ten  times faster,  so  it ought \nto  be  easier  to  get  accurate  numbers  for  them.  The  information  measure  might \nnaturally  be  applied  to  study,  for  example,  the  changes  in  information  content  of \nvisual  cortical  cells  as  a  visual  stimulus is  blurred or  dimmed. \n\nSupported by  NIMH  grant  MH46823 \n\nReferences \n\nGothard,  K.  M.,  Skaggs,  W.  E.,  McNaughton,  B.  L.,  Barnes,  C.  A.,  and  Youngs, \nS.  P.  (1992).  Place  field  specificity  depends  on  proximity of visual  cues.  Soc \nNeurosci  Abstr,  18:1216.  508.10 . \n\nMarkus,  E.  J.,  Barnes,  C.  A.,  McNaughton,  B.  L.,  Gladden,  V.,  Abel,  T.  W.,  and \nSkaggs,  W.  E.  (1992).  Decrease  in  the  information  content  of hippocampal \ncal  cell  spatial  firing  patterns  in  the  dark.  Soc  Neuroscience  Abstr,  18:1216. \n508.12. \n\nMcNaughton,  B.  L.,  Leonard,  B.,  and  Chen,  L.  (1989b).  Cortical-hippocampal \ninteractions  and  cognitive  mapping:  A  hypothesis  based  on  reintegration  of \nthe  parietal and  inferotemporal pathways for  visual processing.  Psychobiology, \n17:230-235. \n\nMcNaughton,  B.  L.,  Barnes,  C.  A.,  Meltzer,  J.,  and  Sutherland,  R.  J.  (1989a). \nHippocampal granule cells  are necessary  for  normal spatial learning but not for \nspatially selective  pyramidal cell  discharge.  Exp  Brain  Res,  76:485-496. \n\nQuirk,  G.  J.,  Muller,  R.  U.,  and  Kubie,  J.  L.  (1990).  The  firing  of hippocampal \nplace  cells  in  the  dark  depends  on  the  rat's  previous  experience.  J  N eurosci, \n10:2008-2017. \n\nRichmond,  B.  J.  and Optican,  L.  M.  (1990).  Temporal encoding of two-dimensional \n\npatterns by  single units in  primate primary visual  cortex:  Ii information trans(cid:173)\nmission.  J  Neurophysiol, 64:370-380. \n\n\f\f", "award": [], "sourceid": 671, "authors": [{"given_name": "William", "family_name": "Skaggs", "institution": null}, {"given_name": "Bruce", "family_name": "McNaughton", "institution": null}, {"given_name": "Katalin", "family_name": "Gothard", "institution": null}]}