One of the more important issues regarding receptive fields is their ability to be spatially selective.  One way that is conceived is the range of spatial frequencies that the cell is sensitive to.  Often a sine wave grating is used as the basic stimulus for examining sensitivity to spatial frequencies.  Below is a first attempt to describe this feature of ganglion cells of the fovea.  Below are the average responses across cells with a given spatial configuration (you can manipulated this configuration).  The inputs are sine wave gratings that vary X times across the input field with the X being the number at the base of the bar.  The for the bar with 8 at its base, it varies 8 times across the input field of 91 receptors.  The 45 represents the nyquist or theoretical limit for sampling for a field with this number of cells.

You can manipulate the cells and see what happens to the output.  For each center with, there ought to be a frequency which generates the strongest input.   I have plotted two different ways to represent spatial selectivity.  The right hand y-axis shows the average responses across the cells.  This way of looking at the cells shows selectivity if the cells respond to some spatial frequencies and not others regardless of the position of the grating with respect to the cell.  The left hand y-axis shows the range of the responses across all of the cells.  If the cells are not positionally insensitive, some will show a lot of responding and other cells not, depending on how the grating falls on the cell and then the range of responses will be high only for some spatial frequencies.  

Since the two sets are bars are on different axes, please watch the scale carefully.

As you can see, the range shows a much greater selectivity for spatial frequency than the average.  This outcome suggests that the cells are very positionally sensitive to the grating.  By that I mean, how the cell falls on the grating will affect the cells response.  That is not terribly surprising for x cells.  but what this means is that individual cells indicate the phase of the grating and not just spatial frequency.  However, this is not to say that cells build an image of the grating.  Below is a graph of the output of these cells for a selections of these grating using the parameters initially on this page (the inhibitory region with a standard deviation of 3 and the excitatory regions with a standard deviation of 1).

Looking at these outputs, the responses to the 2 c/sampling region grating is fairly well reproduced but at a very small amplitude.  The best response is to the 8 c/sampling region grating an it seems to be missing a peak and trough and shows a beat-like pattern in the output.  The range of responses is nearly as strong to the 16 c/sampling region grating but it shows only 2 cycles on the cells.  Thus, the output from these cells is not a clean reproduction of a grating.  

The beat-like pattern is due more to the spacing of the cells relative to the spatial frequency.  Of course, I could arrange the cells so that their spacing matched the spatial frequency and then all the peaks would show and the input to the brain would better match the input grating.  However, it seems greatly unnecessary to space the cells in the this manner when the grating can be recovered from the information as is.  In fact, it seems that it would be a great waste effort to space all the cells selective to all the different ranges of spatial frequencies just so reproduce all of these sizes.  In other words, a little random placement of the cells in the retina will not affect the quality of the output to the brain.