Neural coding of quality of complex olfactory stimuli in lobsters.

1. Extracellular responses to complex biologically relevant stimuli were recorded from 30 primary olfactory cells from excised antennules of spiny lobsters. The stimulus types were natural extracts of crab, mullet, oyster, and shrimp and artificial mixtures of crab, mullet, oyster and shrimp based on the chemical composition of the related extracts. All stimulus types were presented at the following three concentrations: 0.005, 0.05, and 0.5 mM. 2. The responses were expressed as number of spikes per 5 s. Response magnitude increased significantly as a function of concentration. It was significantly greater for the natural extracts than for the related artificial mixtures but was not significantly different among stimulus types within either natural extracts or artificial mixtures. 3. The cells were broadly tuned to all stimuli. Tuning slightly, but significantly, broadened as a function of stimulus concentration. 4. Multidimensional scaling (MDS) was used to evaluate similarities and dissimilarities among stimuli based on population responses. The artificial mixtures and the natural extracts were analyzed separately. Dimensionality of spatial configuration was based on the following three criteria: stress values, squared correlation values, and relevance to quality coding. 5. When applied to the original data, MDS distributed the stimuli in a two-dimensional space where the location of each stimulus was based mainly on stimulus concentration. The results of a simple standardization procedure showed that this distribution resulted mostly from the significant effect of concentration on one of the two features of population responses, which is the absolute magnitude. This standardization procedure equalized the three concentrations in terms of absolute magnitude of evoked response. Consequently, the neural population responses of the 12 stimuli (4 types X 3 concentrations) could be compared based only on their across-neuron patterns (ANPs) (relative amount of activity across neurons). 6. When stress and squared correlation values were used as criteria for dimensionality, the configuration of the artificial mixtures space was best derived from dimensions 1, 2, and 3 of the three-dimensional resolution. When relevance to quality coding was used, the configuration of the artificial mixtures space was best derived from dimensions 1, 3, and 4 of the four-dimensional resolution. Whether stress and squared correlation values or relevance to quality coding were used, the four types of stimuli occupied nonoverlapping spaces.(ABSTRACT TRUNCATED AT 400 WORDS)