A Conceptual and Biological Framework for Primary Visual Perception

Our basic hypothesis is that primary visual perception (PVP) depends upon both intact early visual areas and intact specific high-order parietal areas. However, to make this hypothesis more precise, to distinguish it from related models, and to make it amenable for critical testing require more elaboration. This is especially so given the numerous controversies already extant in this field. Moreover, many discussions of perception have been hindered by multiple definitions of perceptual experience and difficulties distinguishing perceptual phenomena from high-level cognitive processes. Consequently, we seek to identify those neural networks that subserve PVP...

There is currently an insurmountable gap between identifying the neural representations that correlate with perceptual experience and for accounting for such experience itself. If we want to narrow that gap, we may attempt to identify the minimal necessary requirements for perceptual experience because then we will have identified the minimal gap that needs to be closed. We see our reductionistic approach to PVP as itself a novel one that complements the more general approach to the "neural correlates of consciousness" (Crick and Koch 1995) and the global workspace and neuronal workspace models of Baars (1989) and Changeux and Dehaene (2008), respectively.

As a first step toward our objectives, we identify those cortical areas and their subcortical dependencies required for representations of our most basic visual percepts within a self-referential context. This is to suggest that image representations are linked to neural representations of at least a minimal self with a sense of perceptual ownership in order to establish a personal experience of the images and underlying objects.

However, the applicability of this framework extends beyond the obvious linkage of self and world by attention. For example, functions that build upon attentional selection such as intention and decision making also require a link to self either for ownership of the process and/or for its execution. The present work builds upon many pertinent advances in the basic neurobiological, neurological, neuroanatomic, and cognitive neuroscientific literatures as most relevant to the issues at hand. We intend that this conceptual framework for PVP will facilitate interactions between the above disciplines and with neuropsychology and philosophy alike and also be of interest to those who share in the quest for the basis of perceptual experience...

Recursive Loops

Theorists generally have assumed, supported by the above anatomic evidence, that afferent and recurrent streams of neuronal activity seek mutually reinforcing matches within sequences of cortical areas (Milner 1974; Grossberg 1976; Edelman 1978; Koch 1987; Damasio 1989; Mumford 1992; Ullman 1995; Pollen 1999). Such neural activity may be either stimulus specific or task dependent and likely employs both recurrent corticocortical and cortico-pulvinar--cortical projections (Shipp 2004). Even a single bottom-up top-down cycle of neural activity may remove substantial ambiguities in object recognition almost simultaneously at multiple cortical levels (Ephshtein et al. 2008).

There are abundant reasons to suggest that recurrent projections within both the ventral and dorsal loops participate in perceptual experience. For review, please see Pollen (1999). One of the proposals that seems indispensable for reconciling perceptual experience and its dependence on figure-ground segregation as a precondition for object recognition within the ventral loop is that of Grossberg (1994) and Mumford (1994). These authors suggest that figure-ground separation and object recognition cannot progress in the simple bottom-up serial fashion but must occur iteratively within recursive loops.

Minimal Self

We now attempt to relate the normal aspects of processing described above to an associated sense of self as one requisite for the emergence of PVP. Neisser (1988) defined the most primitive sense of self in relationship to the immediate physical environment as the "ecological self." Neisser’s qualifier was motivated by Gibson’s development of "ecological optics" Gibson (1961) specifying how an animal registers the information about the environment inherent in both stationary and moving optic arrays. Gibson (1979) also proposed that perceiving requires the "coperceiving of self."

Gallagher (2000) incorporates Neisser’s ecological self within his own broader definition of a prereflective prelinguistic "minimal self." Gallagher’s minimal self is implicit accepting the givenness of a first-person perspective. Such a perspective is never "a view from nowhere" but rather is always defined "by its relationship to the perceiver’s body which concerns not simply location and posture but action in pragmatic contexts. . ." Gallagher (2006).

Neural Correlates of Perceptual Experience

All current models generally assume that each subjective phenomenal experience is a consequence of a unique neural state (Crick and Koch 1995). This conclusion is consistent with the present model, that of Tononi and Edelman (1998) and Tononi (2008) on ‘‘consciousness as integrated information,’’ the ‘‘neural stance’’ model of Lamme (2006), and the ‘‘central representation’’ model of Taylor (2001, 2007). These models also assume hierarchical and recurrent processing. See also Block (2009).

However, in other respects, these models derive from different sets of primary principles. Our hypothesis differs from other current models primarily in its objective to define the anatomical and functional correlates for the minimally definable entity of PVP.

Tononi (2008) builds upon a very abstract assumption of consciousness as highly integrated information and proceeds to elaborate models to quantitate the information content of higher perceptual functions based upon this assumption. If future such models become increasingly related to functionally entities, it will be of interest to learn whether and how their predictions will converge or diverge with those of the present approach.

The view of Lamme (2006) that recurrent processing either is equivalent to consciousness or a surrogate marker for it is consistent with but goes beyond my earlier proposal (Pollen 1999) that dynamic incongruities must be resolved both within and between ascending and descending streams as a condition for the emergence of perceptual experience. I did not equate recurrent processing in itself with consciousness but rather considered recurrent processing as a means to achieve a consensus of neuronal activity across ascending and descending pathways.

Our primary objective here differs from that of global workspace models (Baars 1989; Changeux and Dehaene 2008) in one major respect. We do not doubt that neuronal activity in the posterior perceptual core can gain entry to the global workspace and, in turn, be influenced by it. However, we believe that the evidence reviewed here suggests that selective activation within the posterior core, in itself, is sufficient to engender primary visual experience. If we are correct, then the quest for the basis for the emergence of PVP reduces to consideration of a substantially smaller set of anatomic and functional entities than is encompassed by the entire global workspace...

Conclusions

We have attempted to define the most basic level of our subjective visual experiences and in identifying their anatomical neural correlates and related functional properties. The model is consistent with the deficits in PVP after injury to V1/ V2 and the specific visual agnosias that occur after lesions of higher extrastriate areas in the ventral loop. The model is also consistent with the known sources of spatial attention and intention in the dorsal loop and the corresponding contralateral inattentional hemispatial neglect syndromes with respect to specific egocentric reference frames after inferior parietal lobe injury. Above all, our model appears to accommodate within in a single concise conceptual framework, an appreciable range of experimental observations arising from diverse neuroscientific disciplines.

The present model also provides a challenge for future studies to determine whether any additional anatomic and functional entities beyond the posterior perceptual core are required for the emergence of PVP. Our analysis has not identified the mechanism by which such subjective experience itself emerges but has defined plausible neural substrata and functional conditions required for such emergence and thus has drawn increased focus to accounting for PVP itself. In identifying the minimal requirements for PVP or at least something close to such requirements—we have narrowed the gap that needs to be closed between identifying the requirements for perceptual experience and for accounting for such experience itself.

("On the Emergence of Primary Visual Perception," Cerebral Cortex September 2011;21, pp.1941-53)

Pollen himself as early as 1999, had questioned Libet's interpretation of the latencies.

Evidence that a sufficiently long duration of steady state activity is an essential requirement for the emergence of phenomenal experience has heen provided Libet (1991), who has shown that trains of repetitive stimuli which engender phenomenal experience require a longer duration than those that evoke unconscious detection above chance levels. Although earlier work by Libet (1964), based in part upon direct electrical stimulation of the human somatosensorv cortex near threshold, suggested that several hundreds of milliseconds of repetitive stimulation was required before phenomenal experience was achieved, other work on direct electrical stimulation of the feline striate cortex (Pollen, 1977) has shown that most striate neurons are silent during the first several hundred milliseconds of stimulation. Thus, the extremely long latencies originally suggested by Libet may be attributed to inhibitory processes prior to the activation of those neuronal assemblies that engender percepts. Even so, his major conclusion regarding the longer stimulus durations required to engender phenomenal experience compared to those for unconscious detection seems secure.

Whether the longer lime duration required for the emergence of the phenomenal state is itself the key requirement for the emergence of visual experience or is a marker for more intense levels of activity within some neuron, neuronal component or local circuit as a consequence of the greater opportunities for temporal summation that accompany such steady conditions remains an open question. In either case, the fundamental mechanism by which these changes may be achieved may be multiplicative or nonlinear effects on the control of contrast gain of afferent activity by efferent activity as has been recently demonstrated for the effects of V1 on LGN neurons (Przybyszewski et al. 1998). If such a similar mechanism holds at cortical levels, then the next steps at further discerning the necessary neural correlates of phenomenal experience may depend upon resolving what the differences are between states accompanying phenomenal experience and those that are subthreshold for it within individual cortical modules. It is not yet obvious that it will be possible to define sufficient conditions for phenomenal experience unless and until we achieve an understanding of such perception itself.

("On the Neural Correlates of Visual Perception," Cerebral Cortex Jan/Feb 1999;9, pp.4-19)

In 2004, Pockett wrote an article in Consciousness and Cognition on "the death of 'subjective backwards referral'." She said that Pollen's "evidence so seriously undermines the data interpretation underlying the notion of subjective backwards referral that it may well have finally buried it."

In the same issue of Consciousness and Cognition, Pollen commented on those who agreed with his analysis of Libet's 1964 paper,

Thus, I believe that our results are consistent with the insights of those who have long suspected a prolonged integrative mechanism and a delayed neuronal activation following threshold direct cortical stimulation Churchland (1981a, 1981b), Glynn (1990, 1991), Gomes (1998, 2002), and especially Pockett (2002), who correctly surmised the existence of a dynamic intratrain facilatory process that we have confirmed at the single cell level. The relationship between intratrain facilitation and a prolonged latency at liminal currents is straightforward. As long as such facilitation is present, it will always be possible to excite a neuron at a lower current than would be required to elicit an action potential with the very first stimulus but only after a sufficient delay for the threshold to be reached.

Pockett (2002) also correctly surmised that no "antedating of conscious experience" is necessary to account for the result of Libet (1993) in the specific case where a peripheral stimulus is experienced before a cortically activated experience via the opposite hemisphere because the cortical activation of adequate neuronal activity is simply delayed. Our results, of course, provide no direct evidence against Libet’s hypothesis for the existence of a subjective back-referral of the onset of conscious experience to the onset of a stimulus in other experimental situations or in natural vision. What our studies do show in this respect is simply that analysis of this issue must take into account that the onset of neuronal activity after threshold direct cortical stimulation for a long train cannot be assumed to begin with the onset of surface stimulation...

Our human results also suggest that the long window of 350–500 ms before conscious experience that appears after threshold cortical stimulation of the striate cortex is also due to the prolonged latency for activation of the underlying cortical neurons. Thus, these results are not indicative of any excessively long central processing time for the emergence of conscious experience after direct cortical stimulation as Libet (1964, 1973, 1993, 2003) has proposed.