- "The proposal (previous discussion - Harris & Arabzadeh) that there is a functional meaning to the preservation of directional selectivity for the onset deflection and loss of directional selectivity for subsequent higher-frequency vibrations is intriguing and makes intuitive sense."

We think that this sort of modulation in time could be highly useful to texture discrimination, and there is circumstantial evidence that it could occur.

- "A second issue concerns stimulus waveform shape... Do the authors have any additional simulations, or perhaps just an intuition, about how directional selectivity would be affected by smooth, continuously changing whisker motion, or for discrete stimuli but with smoother onsets?"

Indeed, in the model we do not test how dependent selectivity is on waveform shape. Having fixed-amplitude deflections on a grid meant that all stimuli were sudden, with equal velocity/acceleration as well as equal amplitude. This allowed us to avoid having to define tuning curves for position or velocity – the only tuning in the model is directional tuning (Fig. 1). We did this purposely to simplify interpretation and keep our assumptions to a minimum, thus isolating the effect of frequency on directional tuning and avoiding “tricks” or artifacts due to interactions with other stimulus variables. This means that with the model in its present form we cannot make predictions for smooth, continuous whisker motion or other stimuli.

However, we do have an experimentally-based intuition for how directional selectivity would be affected by other stimulus waveforms. Here I present the experimental data more fully than in the Discussion, where the argument is only sketched: (1) Directional selectivity is partly based on latency tuning (Wilent and Contreras, 2005), which relies on thalamic synchronicity. This is because the only way for latency tuning with a precision of ~<1 ms to be successful is if thalamocortical inputs get activated with at least that level of precision. If, e.g., an inhibitory gate shuts out excitation that arrives more than, say, 2 ms after the first inputs, then a substantial fraction of the inputs coming from the preferred direction must have arrived within that period. Therefore, inputs coming from a given direction should have very similar timing, i.e. significant synchronicity. (2) However, thalamic synchronicity is stimulus-dependent. Synchronicity is highest for step or ramp-and-hold stimuli (Pinto, Brumberg and Simons, 2000; Temereanca and Simons, 2003; Lee and Simons, 2004). Recent data, however, imply that synchronicity will degrade significantly for less-sharp or continuously varying waveforms, e.g., sinusoids (elegant use of this is made in Bruno and Sakmann, 2006). In an unpublished collaboration with Rasmus Petersen’s group (CNS meeting, 2006; Cosyne meeting, 2007), we have collected responses to white noise stimuli in a population of VPM thalamic neurons. Analyzing stimulus-response relationships with reverse correlation methods, we’ve found that although each neuron’s responses are highly temporally precise and informative, different neurons tend not to fire at uniform times; instead, they tend to respond to slightly different stimulus “features” and fire at different phases in the waveform. This implies that although cells will indeed tend to fire simultaneously when they are probed with a strong, sudden stimulus, the tendency to fire at the same time will depend sensitively on the nature of the stimulus.

Our intuition is that therefore, latency tuning requires fairly sharply rising stimuli, because it requires thalamic synchronicity, which is only high with sharp stimuli. Since latency tuning contributes significantly to directional selectivity, the result should be that smooth, continuous stimuli will cause a decrease in directional selectivity. It would be very interesting to have direct experimental evidence for this idea.