From Shepherd's Synaptic Organization of the Brain (5th ed, pg 513):
In the cat visual cortex, the terminal arbors of each individual
thalamic afferent may extend over 1-5 mm of the cortical surface (Fig. 12.11)
so that each point in layer 4 is covered by the arbors of at least 1000
separate thalamic relay cells. Thus, the dendritic tree
of an average layer 4 neuron, which extends for 200-300 micron, could
receive input from many more thalamic afferents. However, the
connections are not made randomly between the geniculate afferents
and the cortical neurons. Selectivity is expressed in several ways.
For example, there is a high degree of precision in the visuotopic map
recorded in the first-order cortical neurons in the input layer, i.e.,
those receiving mono-synaptic activation by the thalamic afferents.
This clustering is made according to the eye preference of the arbors.
The afferents of those thalamic relay neurons that are driven by the
right eye cluster together in regions about 0.5 mm in diameter and are
partially segregated from the afferents that are driven by the left
eye. This segregation forms the basis of ocular dominance columns.
There are, of course, many other factors in the specificity of thalamocortical projections.
Point here being that thalamocortical projections are not necessarily limited to cortical neurons with a shared receptive field in visual space, but may be limited in other ways.
Interestingly, the corticothalamic feedback projection is also not limited to shared spatial receptive fields (Shepherd, p. 321ff.):
Thus, for the lateral geniculate nucleus, this cortical pathway comes
from visual cortex (mostly areas 17, 18, and 19), and likewise,
somatosensory and auditory cortex project back, re- spectively, to the
ventral posterior lateral and medial geniculate nuclei. One
implication of this reciprocity is that the corticothalamic pathway
faithfully ad- heres to the map established in the thalamic nucleus.
For instance, the corticogenicu- late pathway conforms to the
retinotopic map in the lateral geniculate nucleus. However, there is
some question as to the extent to which the maps match at the cellular
level. This is based on evidence that, in the cat (Murphy and Sillito,
1996), the spread of an individual corticogeniculate axon arbor can be
quite extensive, reaching well beyond the region within which
receptive fields that match those of the cortical axon can be
recorded. The corticogeniculate terminals have a maximal extent of 1.5
mm compared with the spread of a typical retinogeniculate arbor of
only about 0.2-0.4 mm (Bowling and Michael, 1984; Sur et al., 1987).
The retinogeniculate arbor's expanse roughly corresponds to the size
of a geniculate receptive field, implying that the corticogeniculate
axonal arbor can contribute to subtle effects on relay responses
beyond the "clas- sic" receptive field. However, the majority of the
corticothalamic terminals lie in a central core that roughly
corresponds to the classical receptive field.