1.3 The functionally organised visual system
1.3.3 The cortical hierarchy
The cortical areas of the visual system are typically represented as a hierarchy of intercon-
nected areas (Lennie, 1998; Livingstone & Hubel, 1988) (see Figure 1.3). The LGN acts as
the relay centre for the visual pathway. The visual system is broadly divided into multiple
visual areas (Zeki, 1978). Each area is comprised of highly specialised receptive fields and
Receptive fields of a neuron
Receptive fields of a neuron represent a defined region in the visual field which elicits a
response when stimulated (Sincich & Horton, 2005). The receptive fields at the lowest level
areas, V1 and V2, represent simple features over a small area of the visual field and are then
passed to the higher cortical areas such as V3, V4 and V5. The receptive fields of these areas
are larger and pool the sensory information across several lower level receptive fields (Burr
& Thompson, 2011; Gilbert et al., 2001). The strength of response of the neuron depends on
how well the stimulus is matched to its receptive field. In early areas, receptive fields are
tuned to simple properties such as orientation and scale. Cells in these areas are also tuned to
properties such as direction of motion and colour. Receptive fields in higher areas tend to
be larger, more specialised, more broadly tuned to dimensions such as scale, and respond to
more complex features.
The primary visual cortex
The first cortical visual processing level is the primary visual cortex (V1). It is the largest of
all visual areas of the cerebral cortex in the macaque (Felleman & Van Essen, 1991), and is
comprised of six sub-layers (see Figure 1.3). V1 is retinotopically organised and incoming
projections from the LGN are organised by the eye of origin. V1 extracts information of ocular
origin, orientation tuning and visual space and organises these in columnar functional maps
(Callaway & Nassi, 2009). The receptive fields in V1 are small and only respond to a very
restricted area of the visual field (Furlan & Smith, 2016; Lamme, 2003; Simoncelli & Heeger,
fields to build elongated receptive fields. The receptive fields resemble Gabor functions
with odd-symmetric or even-symmetric arrangements of on- and off-regions (Figure 1.4(b)).
These cells represent simple visual dimensions, such as the position, orientation and scale of
local image features (Ahissar & Hochstein, 1997). The specificity of cells in primary visual
cortex was identified using micro electrodes to record the activity of cells in cats (Hubel
& Wiesel, 1959). The position and orientation of the stimulus was crucial in obtaining a
strong reading from a specific neuron. Physiological evidence from a range of animals has
shown that neurons in the primary visual cortex are tuned to orientation (DeValois et al.,
1982; Hubel & Wiesel, 1962), spatial frequency (DeValois et al., 1982; Movshon et al., 1978)
and direction (Livingstone & Hubel, 1988; Movshon & Newsome, 1996). Consequently,
this has also been supported by electrophysiological findings, from the study of area V1 in
monkeys (Gilbert et al., 2001) and functional magnetic resource imaging (fMRI) in humans
(Seitz et al., 2005).
Figure. 1.4 (a) Simple cells in V1 sum the inputs from an array of LGN receptive fields that resemble an elongated bar with an on-centre and off-surround. (b) These receptive fields can be modelled using odd-symmetric and even-symmetric Gabor functions fields
Area V2 is the second largest visual area in the macaque, and receives two thirds of
its input from V1 (Sincich et al., 2003). V2 plays an important role in the processing of
orientation, colour, depth (Sincich & Horton, 2005) and motion (Hu et al., 2018). Recently,
area V2 has also been shown to contribute to binocular depth perception (Hu et al., 2018).
Psychophysical experiments are also used to investigate the behavioural responses of
receptive fields. Local visual tasks are defined as those that can be performed reliably based
on the type of information which individual cells in early cortical areas such as V1 encode.
These include the detection of stimuli and the discrimination of their orientation, position
and direction of motion.
Extrastriate Visual Areas
The higher levels in the visual pathway, such as V3, V4 and V5, have neurons with much
larger receptive fields. The tuning properties of these neurons are more complex, and their
responses are less dependent on the location and retinal size of stimuli, the viewpoint of
the observer and the prevailing lighting conditions (Felleman & Van Essen, 1987; Furlan
& Smith, 2016; Hubel & Wiesel, 1965; Mikami et al., 1986; Movshon et al., 1978; Sillito
et al., 2006; Zeki, 1974). The receptive fields of neurons in higher cortical areas integrate
information to represent global stimulus properties (Amano et al., 2009; Bex & Dakin,
2002; Burr & Thompson, 2011; Gilbert et al., 2001; Nishida, 2011) across a collection
of V1 receptive fields. At each step up the hierarchy the receptive cells get progressively
more complex in their tuning. Higher cortical areas, beyond V4, begin to display category
preferentially to faces (Blake & Wilson, 2011). Furthermore, object-specific cells have been
identified in the lateral occipital cortex (Grill-Spector et al., 2001).
In psychophysical and physiological studies, global processes are investigated using
stimuli or tasks that can only be resolved through integration and segregation of coherent
or conflicting information (Burr & Thompson, 2011; Garcia et al., 2013; Nishida, 2011). A
canonical example of this is the global motion stimuli introduced by Williams and Sekuler
(1984). These stimuli contain dots moving in multiple directions, and the observer’s task is to
determine the direction of motion of the stimulus as a whole. Information in local regions of
the image (and thus also the responses of V1 neurons) will reflect the direction of motion of
individual dots, and the task can only reliably be completed by integrating information about
the directions of all the dots across space. Cortical cells higher in the processing hierarchy
are involved in the perception of global aspects of a percept and generalise across individual
features such as spatial frequency, speed, motion and orientation. Receptive cells in the
global processing levels pool sensory information from a selection of lower level receptive
cells (see Figure 1.5).