As anyone whose nerves have been jangled by a baby's howl or who havebeen riveted by the sight of an attractive person knows, nature hasevolved sensory systems to be exquisitely tuned to relevant input. Amajor question in neurobiology is how neurons tune the strength oftheir interconnections to optimally respond to such inputs.
Neuronal circuitry consists of a web of neurons, each triggeringothers by launching bursts of neurotransmitters at targets on receivingneurons to produce nerve impulses in those targets. Neurons adjust thestrength of those connections adaptively, to amplify or suppressconnections. Some four decades ago, a general principle called the"efficient coding hypothesis" was formulated, holding that sensorysystems adjust to efficiently represent the complex, dynamic sounds,sights, and other sensory input from the environment.
Writing in the August 4, 2005, issue of Neuron,researchers led by Christian K. Machens of Cold Spring HarborLaboratory and Andreas Herz of Humboldt-University Berlin describeexperiments with grasshopper auditory neurons that reveal new detailsof such sensory coding. Their findings show that "optimal stimulusensembles" that trigger the neurons differ from those the grasshopperhears in the natural environment but largely overlap with components ofnatural sounds found in mating and mate-location calls.
In their experiments, the researchers first played varioussnippets of white noise to isolated grasshopper auditory nerves andmeasured the electrophysiological signals reflecting the reactions ofthe auditory neurons to those sounds. These experiments revealed thedistribution of stimuli called the "optimal stimulus ensemble" (OSE)that allowed the neurons in the system to perform optimally.
Once the researchers had characterized the OSE, they thenanalyzed how this measure compared to the neuronal response to naturalsounds--including environmental sounds like the rustling of grass andinsect communication signals such as grasshopper or cricket matingcalls.
They found that the OSEs of the receptors particularly matchedcharacteristic features of species-specific acoustic communicationsignals used by grasshoppers to attract mating partners.
"Hence, instead of maximizing the average information gainedabout natural stimuli, the receptors appear to maximize the informationgained about specific, but less often occurring aspects of thestimuli," concluded the researchers. "This result suggests that anorganism may seek to distribute its sensory resources according to thebehavioral relevance of the natural important stimuli, rather thanaccording to purely statistical principles.
"For instance, if a few important stimuli within the naturalenvironment need to be encoded with high precision, a large part of asystem's coding capacity could be designated to encode these stimuli.Consequently, it may well be that even small subensembles stronglyinfluence the coding strategy of sensory neurons. In this case, theoptimal stimulus ensemble will not match the ensemble of all naturalstimuli encountered by the particular species."
The researchers also concluded that "We therefore suggest thatthe coding strategy of sensory neurons is not matched to the statisticsof natural stimuli per se, but rather to a weighted ensemble of naturalstimuli, where the different behavioral relevance of stimuli determinestheir relative weight in the ensemble."
Machens, Herz, and their colleagues also concluded that theiranalytical technique could yield broader insight into the evolution ofsensory circuitry.
"Our approach presents a systematic way to uncover potentialmismatches between the statistical properties of the naturalenvironment and the coding strategy of sensory neurons. In turn, thesediscrepancies might improve our understanding of the evolutionarydesign of the specific sensory system," they wrote.
The researchers include Christian K. Machens of Cold Spring HarborLaboratory, Cold Spring Harbor, New York; and Tim Gollisch (Presentaddress: Harvard University, Cambridge, Massachusetts), OlgaKolesnikova, and Andreas V.M. Herz of Humboldt-University Berlin andBernstein Center for Computational Neuroscience, Berlin, Germany. Thiswork was supported in part by a Swartz Foundation Fellowship to C.K.M.;a Boehringer Ingelheim Fellowship to T.G.; and grants from the DFG andthe German Federal Ministry for Education and Research to A.V.M.H.
Machens et al.: "Testing the Efficiency of Sensory Coding withOptimal Stimulus Ensembles." Publishing in Neuron, Vol. 47, pages447-456, August 4, 2005. DOI 10.1016/j.neuron.2005.06.015 www.neuron.org
Cite This Page: