Abstract
Some neurons in area MT of the primate brain have been classified as ‘component direction selective’ (CDS) and others as ‘pattern direction selective’ (PDS) using direction tuning tests with grating and plaid stimuli (Movshon et al., 1985). However, a large proportion (∼35%) of the neurons in these tests are consistently judged ‘unclassed’ (U) because they do not fall clearly into either of the two classification zones. Movshon et al., specifically pointed out that the unclassed neurons should not be regarded as an ‘intermediate’ category because they could simply be the by-product of tests and statistics that lack the sensitivity to clearly distinguish the two classes. Despite this admonition, researchers today often refer to this class as ‘intermediate’ and the role of the unclassed neurons remains a mystery. We decided to investigate the sensitivity issue raised by Movshon et al., in an attempt to understand the nature of the unclassified neurons. Using a model of MT pattern neurons (Perrone, Vision Res., 2004; Perrone & Krauzlis, ECVP, 2007) we examined the factors that have an impact on the classification of model neurons. We found that the speed of the test stimuli can have a significant influence on the classification of the PDS-type model neurons. A 10% difference between the model neuron's tuning speed and the plaid speed was often sufficient to shift the normalized pattern correlation coefficient rating (Smith et al., 2005) from PDS to U. MT direction tuning studies typically do not systematically assess the speed tuning of the neuron and so the likelihood of a speed mismatch (and hence misclassification) is high. We conclude that many ‘unclassified’ MT neurons could in fact be perfectly functioning pattern units and concur with Movshon et al.'s conclusion that the testing procedure could be blurring the boundary between the CDS and PDS types.
Supported by the Marsden Fund Council from Government funding, administered by the Royal Society of New Zealand.