Given the age of the visual system I’m surprised that psychology hasn’t tried to link its architecture to the emergence of personality before. Here are a few thoughts on evidence for two-stream theories of vision.
The evolution of a visual system must confer a benefit to an organism for it to bear the cost of its emergence. The stimulation of photoreceptive cells in this system allow an organism to receive information about its environment from beyond the physical boundaries of its own body. To be of benefit this information must, at some point, be interpreted by the organism. Interpretation of stimuli received from the environment implies a mechanism to respond to a perturbation (else the cost of supporting a visual system would serve no benefit to the organism) and therefore facilitate an interaction with the environment.
Successful interpretation of visual stimuli should therefore allow an organism to orientate to salient information within its environment. The two-visual system framework proposed by Milner and Goodale attempts to explain the interpretation of visual information by an organism (Miller & Goodale, 2006).
Discussion within the field of psychology has long debated the nature of this interpretation by an individual. A constructionist approach to the topic takes vision to be an active process where an individual constructs the attributes of a perceived object in the visual field. The observer, in his perception, infers more about the object than provided purely by the information conveyed by stimulation of photoreceptive cells within the retina. Alternatively Gibson (1979) proposed an ecological approach, that an individual “picks-ups” information about his environment and what the objects that occupy it affords him.
Attempts to reconcile the evidence for both theories and derive a unified theory of vision lead to proposals of two stream theories of vision (Trevarthen, 1968; Schneider, 1969). Ungerleider and Mishkin (1982) found evidence of two distinct visual pathways leading from the occipital cortex in monkeys which they termed the dorsal and ventral streams. The ventral path deals with object identification while the dorsal path deals with object location. Evidence of the ventral stream was demonstrated by lesioning areas corresponding to the inferior temporal cortex which left the animals unable to differentiate between objects or “what” they were. Lesioning of the posterior parietal cortex in the dorsal stream left the animals unable to locate a position during a task or “where” they were.
Milner and Goodale’s standpoint is that the goal of vision is to guide the individual for its benefit. The model they propose still incorporates two separate visual streams emerging from early visual cortex. A dorsal stream that extends to the superior regions of the occipito-parietal cortex and a ventral stream that extends to the inferior regions of the occipito-parietal cortex (Milner, 2012). The function performed by the two streams for the benefit of the individual, not the information derived, is of interest. Therefore they see the dorsal and ventral streams serving vision-for-action and vision-for-perception respectively, as opposed to a “where” or “what” pathway.
The two visual systems differ in their functions. The dorsal stream facilitates vision-for-action allowing the individual to navigate and interact within his environment. The ventral stream facilitates vision-for-perception enabling the recognition and identification of previously encountered – and remembered – items. In Milner and Goodale’s model a ventral pathway, similar to the constructionist approach, items are perceived allocentrically – as they pertain to their environment. In the dorsal stream, somewhat analogous to Gibson’s view, items are perceived egocentrically – as they pertain to the individual for his use or manipulation.
Differing in function, the two streams appear to exhibit different characteristics to each other. In terms of memory, the dorsal stream appears to have little capacity to retain information beyond the time taken to execute the required movement. The ventral stream utilizes its ability to store and recall representations of previously encountered objects and events to identify percepts in the environment. To perform different functions, perception and action vision, requires different frames of reference to interpret the environment. To successfully effect an action the individual must use an egocentric frame of reference to precisely carry out the interaction needed such as grasping, lifting or walking. Alternatively when identifying an object or situation within the environment by comparison with previous experiences a relative or allocentric viewpoint is more versatile. A member of a class of object, say an apple, will not be identical to another apple but with have enough similarities to be deemed as a member of the “class apple” (Norman 2002).
The input to the two systems also affects their parameters. Receiving magnocellular input the dorsal system responds quickly high temporal frequencies and low spatial frequencies. The ventral system receiving both parvocellular and magnocellular input responds to higher spatial frequencies than the dorsal stream (Schiller & Logothetis, 1990). In addition the ventral stream operates most affectively with foveal or parafoveal inputs whereas the dorsal stream in less reliant on the acuity of the image formed on the retina.
If the difference between the two pathway is not in the information they receive from the external environment but how that information is interpreted, the presence of different pathways would hint at a distinct neural response characterizing the responses to different aspects of the visual stimuli.
Neuroimaging evidence for Milner and Goodale’s two visual system
Milner and Goodale were able to inform their hypotheses by working with patient DF who suffers from visual form agnosia. While being unable to identify items visually she can reach for and grasp items in a normal manner. Behavioural studies can provide a double dissociation for vision-for-action and vision-for-perception (Milner, 2012).
Functional magnetic resonance imaging (fMRI) studies of DF by James et al. (2003) show no response to line drawings of objects in the lesion site at lateral occiptal (LO) area where control participants show selective activation. However, DF showed strong selective activation in the fusiform and parapippocampal gyri when presented images of texture or colour. Goodale (2008) notes that as well as fitting well with behavioral observation of DF, the LO plays a “special role in processing the geometrical structure of objects” and the activation found rostral to the legion may have a role in the processing of the material constituents of the object.
To investigate the properties a neuronal population is sensitive to Grill-Spector et al. (1999) used the principle of adaption. Functional magnetic resonance adapation (fMR-A) attempts to overcome the constraints of spatial resolution of fMRI by measuring the change in the magnetic resonance signal of a population after adaption to a stimuli. Within a voxel there may be neurons belonging to many cortical networks that respond to different properties that cannot be resolved by the technique. A fMR-A paradigm repeatedly exposes a participant to a like visual stimuli, adaption causes a reduction in the magnetic resonance signal from the monitored region of interest (ROI). Then by altering a property of the stimulus such as size, translation, illumination or viewpoint and measuring the recovery, if any, of the MR signal an assessment of the population’s ability to code for the altered property can be made. A recovery in the MR signal would show that the neuronal population did code for the altered property as the signal increased when the population was no longer constantly selectively activated. A failure of the MR signal to recover would mean that the population did not code for the altered property as the population was still adapted despite changes in the measure being assessed.
Grill-Spector et al. found that objects in the lateral occipital complex (LOC) were less sensitive to changes in size and position of stimulus, meaning they were not coded for them, than changes in illumination and viewpoint. The experiment shows that, in the case presented, properties of the scene returned to the individual (i.e. changes in size and location) related to the egocentric viewpoint were not found in the proposed ventral stream, supporting Milner and Goodale’s hypotheses.
However a priming experiment recorded using BOLD fMRI by James et al. (2002) found a reduction of activation in the LOC for viewpoint when rotated viewpoints of an object were presented to participants. Without clarification this would be a surprising finding using Milner and Goodale’s hypotheses. One factor the experimenters attribute to a possible explanation is the delay in presenting a rotated viewpoint of the same object was in the order of minutes in their experiment versus an immediate presentation in the experiment of Grill-Spector et al. That is an object was more likely to be seen as “the same thing” when presented from a different viewpoint after several minutes than when presented immediately. The cognitive load of storing egocentric representations in a continuously changing environment, would overload the individual if stored for a prolonged period of time. The delay in presentation by James et al.(2002) may be long enough for egocentric viewpoint representations to decay and the recall of the semantic properties of the item to become prominent to the participant and attribute to an item. This would add support Milner and Goodale’s hypotheses that the dorsal and ventral stream are distinctly different.
James et al. (2002) also found increased activation in the caudal area of the intraparietal sulcus (IPS), upon presentation of rotated images. Their conclusion is that the selective response by the dorsal stream area “treated rotated images as new objects”. In addition the priming of the caudal IPS in the first instance shows that the dorsal stream is capable of responding selectively to the presentation of objects without an associated action. A differentiation of rotated objects from an egocentric view would be in agreement with a vision-for-action standpoint.
To determine whether the response to objects in the dorsal stream was due to rotation of the object or it’s identify, which could not be determined from the work of James et al.(2002), Valyear et al.(2006) employed an event-related design. Being able to isolate factors of identity and orientation they were able to find selective activation for identity within the LOC (near the ventral temporo-occipital junction, concurring with James et al.) with this area not being selective for orientation. They also found selective activation for orientation at the occipito-parietal junction (OPJ), adjacent to the IPS, which they found to be selective for orientation not identity. Valyear et al. claim that their results lend support to the argument that the ventral stream supports perception while the dorsal stream is “more involved in the visual control of actions that might be directed at those objects”, which agrees with Milner and Goodale’s hypotheses.
Konen et al. (2011) used both fMRI and fMR-A techniques in a case study of patient SM. SM acquired object form agnosia and prosopagnosia as a result of a closed head injury in a motor vehicle accident. Structural neuroimaging reveals a lesion adjacent to areas hV4 and VO1 of the right hemisphere. As expected SM showed significantly less selective activation of areas of the LOC in the right fusiform gyrus than a control group when presented with images of objects. Surprisingly similar levels of (in)activation where found in the intact left hemisphere of SM. The experimenters draw a conclusion that the distance effect shows that the “right lateral fusiform gyrus is critically involved in object recognition”.
Unlike the control group, fMR-A time course data between the left and right hemisphere shows no correlation for SM, implying that SM fails to adapt. If true, this may also imply that the ventral stream fails to recognise, as a result of the lesion, a repeatedly presented object as, following James et al. (2002), “the same thing” and therefore form visual memories. This failure would make the ventral stream unable to recall semantic properties of an object, interfering with perceptions from an allocentric viewpoint. This inability combined with the dorsal stream’s short-term capacity for memory may give an insight into visual agnosia and explain how an object can be described by a patient via other sensory pathways when he fails to do so through visual recognition. The inability, as noted by Goodale in patient DF, to describe the geometrical structure of an object combined with the inability of the dorsal stream to provide long-term memory of an object may lead to conditions that would describe object form agnosia.
Evidence for interaction between visual streams
In general usage vision appears continuous and seamless to the observer, despite experimental evidence for two pathways. At what point do the two streams combine? Milner and Goodale argue that “the contributions made by the two streams are quite distinct, and that establishing how they work together is the key to a full understanding of visually guided behavior” (Milner & Goodale, 2010). However strictly defining vision-for-action and vision-for-perception leads two criticism of the model (Jeannerod & Jacob, 2005; Rizzolatti & Matelli, 2003;). Schenk and McIntosh (2010) note this may be resolved if the pathways are not taken to be exclusively and mutually independent of each other. However their contention that “specializations of the two streams are relative, not absolute” are rejected by Milner and Goodale (2010).
Konen and Kastner (2008) applied a fMR-A paradigm to regions of interest in the dorsal stream and found evidence of adaptation of both “size and viewpoint-invariant properties similar to LOC” in the posterior IPS. Evidence of allocentric coding within the dorsal stream is at odds with Milner and Goodale’s claim of two distinct pathways. In addition James et al. (2002) noted that the IPS was selectively activated despite an object having no action associated with it.
Van Polanen and Davare (2015) suggest that interaction between the two the two proposed streams are required for complex tasks. While visual form agnosia patient DF may be able be able to perform simple reaching and grasping actions on par with non-lesioned participants her performance is impinged in more complex tasks where semantic knowledge is required (Carey et al., 1996; Dijkerman et al., 2009)
Ramayya et al. (2010) used diffusion tensor imaging (DTI) to investigate brain networks activated during tool use with the hypothesis that the middle temporal cortex is involved in supplying semantic information in completion of an action based task. Taken without context, the involvement of an area associated with the ventral stream in a task typically thought to be mediated by areas in the dorsal stream would contradict Milner and Goodale’s model.
The researchers found a pathways from the middle temporal gyrus (MTG) to the anterior supramarginal gyrus (aSMG) and from the MTG to the supramarginal gyrus (SMG) and angular gyrus (AG) in the parietal lobe. In addition pathways from both aSMG and SMG/AG were found to the frontal lobe (See Figure 1). Notably the MTG-aSMG pathway was only found in the left hemisphere whereas MTG-SMG/AG pathway was found bilaterally. It is suggested that the unilateral presence of the MTG-aSMG pathway conveys semantic and non-spatial information about the item that is needed to formulate an action based motor solution such as estimations of the item’s weight or fragility. The MTG-SMG pathway is proposed as a pathway to convey learned behaviour appropriate to the visual-motor execution of the task. This dissociates well with Konen et al.‘s finding that the right fusiform gyrus is vital for object recognition, while the supply of information from the ventral stream of the geometrical nature of the object or tool by a pathway in the left hemisphere is needed to add context to an action.
Figure 1. Reproduced from Ramayya et al. (2010)
Conclusion
Neuroimaging studies give broad support to Milner and Goodale’s two visual system hypothesis. fMRI and behavioural studies of visual form agnosia patient DF provides a double dissociation between behavior and selective activation. Areas selectively activated in controls when presented with images of objects overlapped with the lesioned area in DF. In addition, DF could complete simple visual-motor based tasks as well as non-lesioned individuals. This adds weight to the theory that it is the function that the information is used for that guides the individual. However DF was found to underperform on complex visual-motor based tasks as noted by Carey et al..
Milner and Goodale (2010) maintain that establishing how the two visual streams work together is the “key to a full understanding of visually guided behaviour”. Ramayya et al. used DTI to provide evidence of interaction between aspects of cortical networks underlying the two streams. This evidence was shown in the context of the ventral stream providing information to the dorsal stream during tool use. This dovetails with the work of Carey et al. who note that DF made errors when asked to make predictions about tool use.
Milner and Goodale’s hypotheses fails if the streams are strictly linked to cortical areas rather than considering them as operating functions upon information. fMR-A studies by Konen and Kastner (2008) found evidence of allocentric coding in the posterior IPS – an area associated with the dorsal stream. Evidence found in this ROI does not preclude the activation being a result of activation elsewhere being fed forward as suggested by Ramayya et al. In addition investigation the time-course and order of processing of information may lead to more clues about the integration and operation of the two visual streams. Integration of perception and action must occur at some point in the visual stream. Neuroimaging studies may not be sophisticated enough to distinguish between this integration of functional streams or selective action and vision activation within a stream and hence provide disagreement as to whether the two streams are exclusive as maintained by Milner and Goodale.
The technological limitation of current neuroimaging techniques may preclude them from fully supporting Milner and Goodale’s hypotheses or help refute claims against them. A study by Dubois et al. (2015) found that in some instances multi-variate pattern analysis was not able to resolve information that was present in single cell recordings in macaques. Though not in humans, this study is particularly relevant as the MVPA was found to be able to decode information on identity when the animals were presented with images of humans. However the technique could not distinguish between different orientations of the individual, which the neurophysical techniques could. If egocentric or vision-for-action information cannot be resolved by neuroimaging techniques – but is known to be present through others – they can never completely support or reject Milner and Goodale’s hypotheses.
References
Carey, D.P., Harvey, M., Milner, A.D. 1996. Visuomotor sensitivity for shape and orientation in a patient with visual form agnosia. Neuropsychologia, 34, 329–337.
Dijkerman, H.C., McIntosh, R.D., Schindler, I., Nijboer, T.C.W., Milner, A.D. 2009. Choosing between alternative wrist postures: action planning needs perception. Neuropsychologia, 47, 1476–1482.
Dubois, J., de Berker, A. O., & Tsao, D. Y. (2015). Single-unit recordings in the macaque face patch system reveal limitations of fMRI MVPA. The Journal of Neuroscience : The Official Journal of the Society for Neuroscience, 35(6),
James, T. W., Humphrey, G. K., Gati, J. S., Menon, R. S., & Goodale, M. A. (2002). Differential effects of viewpoint on object-driven activation in dorsal and ventral streams.Neuron, 35(4), 793-801.
James, T. W., Culham, J., Humphrey, G. K., Milner, A. D., & Goodale, M. A. (2003). Ventral occipital lesions impair object recognition but not object-directed grasping: A fMRI study. Brain, 126, 2463–2475.
Jeannerod, M., & Jacob, P. (2005). Visual cognition: A new look at the two-visual systems model. Neuropsychologia,
43, 301–312.
Konen, C. S., & Kastner, S. (2008). Two hierarchically organized neural systems for object information in human visual cortex. Nature Neuroscience, 11(2), 224-231.
Konen, C., Behrmann, M., Nishimura, M., & Kastner, S. (2011). The functional neuroanatomy of object agnosia: A case study. Neuron, 71(1), 49-60.
Gibson, J. J. (1979). The ecological approach to visual perception. Boston (etc.); London;: Houghton Mifflin.
Goodale, M. A. (2008). Action without perception in human vision. Cognitive Neuropsychology, 25(7), 891-919.
Grill-Spector, K., Kushnir, T., Edelman, S., Avidan, G., Itzchak, Y., & Malach, R. (1999). Differential processing of objects under various viewing conditions in the human lateral occipital complex. Neuron, 24(1), 187-203.
Norman, J. (2002). Two visual systems and two theories of perception: An attempt to reconcile the constructivist and ecological approaches. Behavioral and Brain Sciences, 25(1), 73-96.
Goodale, M. A., & Milner, A. D. (2010). Two visual streams: Interconnections do not imply duplication of function. Cognitive Neuroscience, 1(1), 65-68.
Milner, A. D. (2012). Is visual processing in the dorsal stream accessible to consciousness? Proceedings: Biological Sciences, 279(1737), 2289-2298.
Milner, A. D., & Goodale, M. A. (2006). The visual brain in action (2nd edition). Oxford: Oxford University Press.
Goodale, M. A., & Milner, A. D. (2010). Two visual streams: Interconnections do not imply duplication of function. Cognitive Neuroscience, 1(1), 65-68.
Ramayya, A. G., Glasser, M. F., & Rilling, J. K. (2010). A DTI investigation of neural substrates supporting tool use. Cerebral Cortex, 20(3), 507-516.
Rizzolatti, G., & Matelli, M. (2003). Two different streams form the dorsal visual system: Anatomy and functions. Experimental Brain Research, 153, 146–157.
Schenk, T., & McIntosh, R. D. (2010). Do we have independent visual streams for perception and action? Cognitive Neuroscience, 1(1), 52-78.
Schiller, P.H., Logothetis, N.K., Charles, E.R. (1990). Role of the color-opponent and broadband channels in vision. Visual Neuroscience, 5, 321–346.
Schneider, G.E., (1969). “Two visual systems” Science, 163 (3870), 895–902.
Trevarthen, C. B. (1968). Two mechanisms of vision in primates. Psychologische Forschung, 31(4), 299.
Ungerleider, L. G., & Mishkin, M., (1982). Contribution of striate inputs to the visuospatial functions of parieto-preoccipital cortex in monkeys. Behavioural Brain Research, 6(1), 57.
Valyear, K. F., Culham, J. C., Sharif, N., Westwood, D., & Goodale, M. A. (2006). A double dissociation between sensitivity to changes in object identity and object orientation in the ventral and dorsal visual streams: A human fMRI study. Neuropsychologia, 44(2), 218-228.
van Polanen, V., & Davare, M. (2015). Interactions between dorsal and ventral streams for controlling skilled grasp. Neuropsychologia, 79(Pt B), 186.