{"title": "Cricket Wind Detection", "book": "Advances in Neural Information Processing Systems", "page_first": 802, "page_last": 807, "abstract": null, "full_text": "802 \n\nCRICKET WIND  DETECTION \n\nJohn  P.  Miller \n\nNeurobiology  Group,  University  of California, \n\nBerkeley,  California  94720,  U.S.A. \n\nA  great  deal  of interest  has  recently  been  focused  on  theories  concerning \n\nparallel  distributed  processing  in  central  nervous  systems.  In  particular, \n\nmany researchers have become very interested in the structure and function \n\nof \"computational maps\"  in sensory  systems.  As  defined  in a  recent review \n\n(Knudsen et al, 1987), a  \"map\" is  an array of nerve cells, within which there \n\nis  a systematic variation in the  \"tuning\" of neighboring cells for  a particular \n\nparameter.  For example, the projection from retina to visual cortex is  a rel(cid:173)\n\natively simple topographic map; each cortical hypercolumn itself contains a \n\nmore complex  \"computational\" map of preferred line orientation represent(cid:173)\n\ning the angle of tilt of a  simple line stimulus. \n\nThe overall goal of the research  in  my  lab is  to  determine  how  a  relatively \n\ncomplex mapped sensory system extracts and encodes information from  ex(cid:173)\n\nternal  stimuli.  The  preparation  we  study  is  the  cercal  sensory  system  of \n\nthe cricket,  Acheta  domesticus.  Crickets  (and many other insects) have two \n\nantenna-like appendages at the rear of their abdomen, covered with hundreds \n\nof \"filiform\"  hairs, resembling bristles on a bottle brush.  Deflection of these \n\nfiliform  hairs  by  wind  currents  activates  mechanosensory  receptors,  which \n\nproject  into the terminal abdominal ganglion to form  a  topographic  repre(cid:173)\n\nsentation (or \"map\") of \"wind space\".  Primary sensory interneurons having \n\n\fCricket Wind Detection \n\n803 \n\ndendritic  branches  within  this  afferent  map  of  wind  space  are  selectively \n\nactivated  by  wind  stimuli  with  \"relevant\"  parameters,  and  generate  action \n\npotentials  at  frequencies  that  depend  upon  the  value  of those  parameters. \n\nThe  \"relevant\"  parameters  are  thought  to  be  the  direction,  velocity,  and \n\nacceleration of wind  currents  directed  at the animal  (Shimozawa &  Kanou, \n\n1984a &  b).  There are only  ten  pairs of these interneurons  which carry the \n\nsystem's output  to higher  centers.  All  ten of these output units  are identi(cid:173)\n\nfied,  and  all  can  be monitored  individually  with  intracellular electrodes  or \n\nsimultaneously  with  extracellular  electrodes.  The  following  specific  ques(cid:173)\n\ntions are currently being addressed:  What are the response properties of the \n\nsensory receptors,  and  what are the  I/O  properties  of the receptor layer as \n\na  whole?  What  are  the  response  properties  of all  the  units  in  the  output \n\nlayer?  Is  all  of the  direction,  velocity  and  acceleration  information  that  is \n\nextracted  at  the  receptor  layer  also  available  at  the  output  layer?  How  is \n\nthat  information  encoded?  Are  any  higher  order  \"features\"  also  encoded? \n\nWhat is  the overall  threshold,  sensitivity and  dynamic range of the system \n\nas  a  whole for  detecting features  of wind stimuli? \n\nMichael  Landolfa is  studying the sensory neurons  which serve as  the inputs \n\nto the cercal  system.  The sensory  cell  layer  consists  of about  1000  afferent \n\nneurons, each of which innervates a single mechanosensory hair on the cerci. \n\nThe input/output relationships of single sensory neurons were characterized \n\nby  recording from  an  afferent  axon  while  presenting appropriate stimuli  to \n\nthe sensory hairs.  The primary results were as follows:  1) Afferents are direc(cid:173)\n\ntionally  sensitive.  Graphs of afferent  response  amplitude versus  wind direc(cid:173)\n\ntion are approximately sinusoidal, with distinct preferred and anti-preferred \n\ndirections.  2)  Afferents  are  velocity  sensitive.  Each  afferent  encodes  wind \n\n\f804 \n\nMiller \n\nvelocity  over  a  range  of approximately  1.5 log  units.  3)  Different  afferents \n\nhave  different  velocity  thresholds.  The overlap  of these different  sensitivity \n\ncurves  insures  that  the  system  as  a  whole  can  encode  wind  velocities  that \n\nspan several log units.  4)  The nature of the afferent  response  to  deflection \n\nof its sensory hair indicates that the parameter transduced by the afferent is \n\nnot hair displacement, but change in  hair displacement.  Thus, a  significant \n\nportion  of the processing  which  occurs  within  the  cereal  sensory  system is \n\naccomplished  at the level of the sensory afferents. \n\nThis information about the direction and velocity of wind stimuli is encoded \n\nby the relative firing  rates of at least  10  pairs of identified sensory interneu(cid:173)\n\nrons.  A  full  analysis of the input/output properties of this system requires \n\nthat the activity of these output neurons be monitored simultaneously.  Shai \n\nGozani has implemented a computer-based system capable of extracting the \n\nfiring  patterns of individual neurons  from  multi-unit  recordings.  For  these \n\nexperiments, extracellular electrodes were arrayed along the abdominal nerve \n\ncord in each  preparation.  Wind stimuli of varying directions, velocities  and \n\nfrequencies  were  presented to the  animals.  The  responses  of the cells  were \n\nanalyzed  by spike descrimination software  based on  an  algorithm originally \n\ndeveloped  by Roberts and Hartline (1975).  The algorithm employs multiple \n\nlinear filters,  and is capable of descriminating spikes  that were coincident in \n\ntime.  The number of spikes  that could  be  descriminated was roughly equal \n\nto  the  number  of independent  electrodes.  These  programs  are  very  pow(cid:173)\n\nerful,  and  may  be of much  more general  utility for  researchers  working  on \n\nother invertebrate and  vertebrate  preparations.  Using  these  programs  and \n\nprotocols,  we  have  characterized  the  output  of  the  cereal  sensory  system \n\nin terms of the  simultaneous activity patterns of several pairs  of identified \n\n\fCricket Wind Detection \n\n805 \n\nin terneurons. \n\nThe results of these multi-unit recording studies, as well  as studies using sin(cid:173)\n\ngle intracellular  electrodes,  have  yielded  information  about  the  directional \n\ntuning and  velocity sensitivity of the first  order sensory interneurons.  Tun(cid:173)\n\ning  curves  representing  interneuron  response  amplitude  versus  wind  direc(cid:173)\n\ntion are approximately sinusoidal,  as  was  the case for  the sensory  afferents. \n\nSensitivity curves representing interneuron  response  amplitude versus  wind \n\nvelocity are sigmoidal, with  \"operating ranges\"  of about  1.5 log units.  The \n\ninterneurons are segregated into several distinct classes having  different but \n\noverlapping  operating  ranges,  such  that  the  direction  and  velocity  of any \n\nwind stimulus can be uniquely represented as the ratio of activity in the dif(cid:173)\n\nferent interneurons.  Thus, the overlap of the different direction and velocity \n\nsensitivity curves in  approximately 20  interneurons insures that the system \n\nas  a  whole  can  encode the characteristics of wind  stimuli  having  directions \n\nthat span 360 degrees and velocities that span at least 4 orders of magnitude. \n\nWe are particularly interested in the mechanisms underlying directional sen(cid:173)\n\nsitivity in some of the first-order  sensory interneurons.  Identified  interneu(cid:173)\n\nrons with different morphologies have very different  directional sensitivities. \n\nThe excitatory receptive fields  of the different interneurons have been shown \n\nto be directly related to the position of their dendrites within the topographic \n\nmap of wind space formed  by the filiform  afferents  discussed  above (Bacon \n\n& Murphey,  1984;  Jacobs & Miller,1985;  Jacobs,  Miller  & Murphey,  1986). \n\nThe precise  shapes  of the directional  tuning  curves  have  been  shown  to be \n\ndependent  upon  two  additional factors.  First, local  inhibitory interneurons \n\ncan  have a  stong influence  over  a  cell's  response  by  shunting excitatory  in(cid:173)\n\nputs from particular directions, and by reducing spontaneous activity during \n\n\f806 \n\nMiller \n\nstimuli from  a cells  \"null\"  direction.  Second, the \"electroanatomy\" of a neu(cid:173)\n\nron's dendritic branches determines the relative weighting of synaptic inputs \n\nonto its different  arborizations. \n\nSome  specific  aims  of our continuing research  are as  follows:  1)  to charac(cid:173)\n\nterize the  distribution of all synaptic inputs onto several different  types of \n\nidentified interneurons, 2) to measure the functional properties of individual \n\ndendrites  of these  cell  types,  3)  to  locate  the  spike  initiating  zones  of the \n\ncells, and 4) to synthesize a  quantitative explanation of signal processing by \n\neach cell.  Steps 1,2 & 3 are being accomplished through electrophysiological \n\nexperiments.  Step  4 is  being accomplished  by developing a  compartmental \n\nmodel  for  each  cell  type  and testing the  model  through further  physiologi(cid:173)\n\ncal experiments.  These computer modeling studies are being carried out by \n\nRocky  Nevin  and John  Tromp.  For these  models,  the structure of each in(cid:173)\n\nterneuron's dendritic branches are of particular functional importance, since \n\nthe flow  of bioelectrical currents through these branches determine how  sig(cid:173)\n\nnals received from \"input\" cells are \"integrated\" and transformed into mean(cid:173)\n\ningful output which  is transmitted to higher centers. \n\nWe  are now  at  a  point  where  we  can  begin  to understand  the operation of \n\nthe system as  a  whole in  terms of the structure, function  and synaptic con(cid:173)\n\nnectivity of the individual neurons.  The  proposed  studies  will  also lay the \n\ntechnical  and  theoretical ground work  for  future  studies  into the  nature of \n\nsignal  \"decoding\"  and  higher-order  processing  in  this  preparation,  mecha(cid:173)\n\nnisms underlying the development, self-organization and regulative plasticity \n\nof units within this computational map, and perhaps information processing \n\nin more  complex mapped sensory systems. \n\n\fCricket Wind Detection \n\n807 \n\nREFERENCES \n\nBacon, J.P. and Murphey, R.K. (1984) Receptive fields of cricket (Acheta do(cid:173)\n\nmesticus) are determined by their dendritic structure.  J.Physiol. (Lond) \n\n352:601 \n\nJacobs, G.A. and Miller, J.P. (1985) Functional properties of individual neu\u00b7 \n\nronal branches isolated  in situ by laser photoinactivation.  ScierJ~, 228: \n\n344-346 \n\nJacobs,  G.A.,  Miller,  J.P.  and  Murphey,  R.K.  (1986)  Cellular  mechanisms \n\nunderlying  directional  sensitivity  of an  identified  sensory  interneuron. \n\nJ.  Neurosci.  6(8):  2298-2311 \n\nKnudsen,  E.I.,  S.  duLac  and  Esterly,  S.D.  (1981)  Computational  maps  in \n\nthe brain.  Annual Review of Neuroscien~ 10;  41-66 \n\nRoberts,  W.M.  and  Hartline,  D.K.  (1915)  Separation  of multi-unit  nerve \n\nimpulse  trains  by  a  multi-channel  linear  filter  algorithm.  Brain  Res. \n\n94:  141- 149. \n\nShimozawa,  T.  and  Kanou,  M.  (1984a)  Varieties  of filiform  hairs:  range \n\nfractionation  by sensory  afi'erents  and  cereal interneurons of a  cricket. \n\nJ.  Compo  Physiol.  A. 155:  485-493 \n\nShimozawa, T. and Kanou, M. (1984b) The aerodynamics and sensory phys(cid:173)\n\niology of range fractionation in the cereal filiform sensilla of the cricket \n\nGryllus bimaculatus.  J.  Compo  Physiol.  A. 155:  495-505 \n\n\f\f", "award": [], "sourceid": 112, "authors": [{"given_name": "John", "family_name": "Miller", "institution": null}]}