If the excitatory input from geniculate relay cells is the dominant input to simple cells and defines their subfields, then many of the response properties of simple cells should resemble those of the relay cells. That both relay cell centers and simple cell subfields come in ON and OFF varieties is one of the resemblances that led Hubel and Wiesel to propose their model, but it is only the most rudimentary resemblance between the two receptive field types. In addition, the widths of simple cell subfields are comparable to those of relay cell receptive field centers at similar visual field eccentricities (Reid and Alonso, 1995, Bullier et al., 1982, Mullikin et al., 1984). The resemblance extends to more subtle measures as well, including the dynamics of the responses to flashing bars (Mullikin et al., 1984) and sinusoidally modulated bars (Saul and Humphrey, 1992), and to the linearity of spatial summation as measured with sinusoidal gratings (Ferster and Jagadeesh, 1991). For each of these measures, simple cell responses fall naturally into the same categories as relay cells, including X and Y, or lagged and non-lagged.
The feedforward model makes an important prediction regarding the relationship between the degree of orientation selectivity and the aspect ratio (length-to-width ratio) of a simple cell's subfields. The longer the subfield in relation to its width, the greater will be the difference in the magnitude of the geniculate excitation evoked by an optimally oriented stimulus and one at right angles. Furthermore, in cells with long narrow subfields, a relatively small shift in stimulus orientation will move a large proportion of the stimulus out of the subfield. Accordingly, the longer a simple cell's subfields are in relation to their width, the more sharply orientation tuned the cell should be. To test these predictions of the feedforward model, (Jones and Palmer, 1987a,b) made high resolution maps of the subfields of simple cells, applying their reverse correlation technique to the responses to small dots flashed briefly throughout the receptive field. The average measured subfield aspect ratio was approximately 5, with values as high as 12 in some cells. Similar values were obtained by Gardner et al. (1999) (but see Pei et al., 1994). More critically, Jones and Palmer and Gardner et al measured orientation tuning in the cells whose receptive fields they had mapped and found a strong relationship between orientation tuning width and receptive field shape: As predicted, the higher the aspect ratio of the subfield, the sharper the orientation tuning. The tuning, however, was sharper than that predicted by the simple feedforward model, a finding that may be related to the effects of the spike threshold (see below).
The feedforward model also predicts that orientation tuning width, when measured with gratings,1should decrease with increasing stimulus spatial frequency (see Troyer et al., 1998). Several studies have found this to be the case (Jones et al., 1987, Hammond and Pomfrett, 1990, Vidyasagar and Sigüenza, 1985, Webster and De Valois, 1985).