How to (and how not to) think about top-down influences on visual perception
Introduction
One influential debate about perception concerns its purity: is perception an encapsulated process that is protected from influences by cognition or is perceptual bottom-up processing influenced by top-down cognitive information? This debate, which is frequently referred to as the ‘cognitive penetration debate’, is complicated by the fact that it is often not clear what kind of mental state is supposed to be doing the penetrating and what kind of mental state is supposed to be penetrated. In other words, it is not clear what is ‘top’ and what is ‘bottom’ in the debate about top-down influences on perception.
In the first part of this paper, we attempt to clarify some of these conceptual issues. We then proceed to suggest a practical alternative to the philosophical turf wars concerning the extent to which perception is encapsulated. We start with two uncontroversial observations: First, even the most avid proponents of the view that perception is cognitively impenetrable accept the existence of top-down processing within the visual system1; in other words, it is uncontroversial that higher-levels of visual processing feed back information to and shape lower-levels of visual processing (Pylyshyn, 1999). And second, the exact locus of the boundary between perception and cognition is notoriously difficult to determine (Bayne, 2009, Kriegel, 2007, Masrour, 2011, Nanay, 2011, Nanay, 2012, Nanay, 2013, Siegel, 2007, Siewert, 2002). On the basis of these two observations, we argue that important insights might be gained once we stop focusing exclusively on top-down modulation of perception by cognition; rather, we suggest that it is heuristically valuable to view this special case within the broader context of top-down influences in a hierarchically organised information processing system.
The general agreement concerning top-down processing within the visual system can be used as a starting point from which our understanding can be expanded to potential higher-level top-down influences without having to commit a priori to what exactly counts as perception or cognition. Once such an approach has been adopted, we can start asking nuanced questions about the specific mechanism of top-down modulation in information processing in general. Among other questions, we can ask what kinds of constraints exist on top-down influences between certain levels of processing and whether some of these constraints might amount to a full-blown boundary as proposed by the encapsulation hypothesis. According to our evaluation of the theoretical and empirical evidence, there is no reason to assume that top-down processing is restricted to specific parts of the information processing hierarchy. By contrast, we defend a view that puts clear constraints on all sorts of top-down processing – as well as on bottom-up processing – but that allows bidirectional flow of information between levels that some would consider to cross the perception-cognition divide.
Section snippets
Two debates about top-down influences on perception
The main conceptual confusion concerning debates about top-down influences on perception is that it is not clear what is meant by ‘perception’ in this context. Many philosophers (Macpherson, 2012, Siegel, 2011, Stokes, 2012), but also some psychologists (e.g., Firestone & Scholl, in press), take ‘perception’ to be perceptual experience: something we are consciously aware of. According to this conceptualisation, the question is whether top-down influences can alter the way we experience a scene
Cognitive penetration and top-down processing
We view the question of cognitive penetration as part of the wider research endeavour that attempts to understand top-down influences on perception. We believe that this perspective has various advantages over an exclusive focus on cognitive influences on perceptual processing. The most important benefit of such an approach is that it provides us with a framework that allows us to ask more nuanced questions: the question of top-down influences on perception is no longer a simple yes-no question
Top-down processing within the visual system
In the past decades, our understanding of the structural and functional organisation of visual systems in various mammalian species has improved dramatically and, in nonhuman and human primates, an increasing number of cortical areas dedicated to processing visual inputs has been and still is being characterised (for reviews see Bullier, 2004, Felleman and Van Essen, 1991, Grill-Spector and Malach, 2004, Katzner and Weigelt, 2013, Van Essen, 2004, Van Essen and Maunsell, 1983). In humans and
Top-down influences from outside the visual system
In recent years, several studies provided evidence to suggest that processing levels that most would consider to be outside the visual system can exert top-down influences on early vision. Here, we will discuss examples of how expectation and memory representations influence processing at early visual stages and we will argue that the function of such top-down effects can be readily understood in terms of the well-established processes discussed in the previous section.
Philosophers,
Conclusions
Humans and nonhuman primates live in a highly complex and ever-changing visual world. In order to deal with the resultant uncertainty and to generate adaptive representations that can guide successful behaviour, visual information processing is highly interactive with an important role for top-down influences from higher-level representations onto lower-level perceptual processing. While some of these higher-level representations can safely be considered to be perceptual – and the resultant
Acknowledgements
This work was supported by the FWO Odysseus grant G.0020.12N and the FWO Research grant G0C7416N.
References (89)
- et al.
Contribution of feedforward, lateral and feedback connections to the classical receptive field center and extra-classical receptive field surround of primate V1 neurons
Progress in Brain Research
(2006) - et al.
The “independent components” of natural scenes are edge filters
Vision Research
(1997) An information-processing analysis of the functional architecture of the primate neocortex
Journal of Theoretical Biology
(2004)Integrated model of visual processing
Brain Research Reviews
(2001)Visual cognition
Vision Research
(2011)- et al.
The development of implicit intergroup cognition
Trends in Cognitive Sciences
(2008) - et al.
Brain states: Top-down influences in sensory processing
Neuron
(2007) - et al.
Visual cortical networks: Of mice and men
Current Opinion in Neurobiology
(2013) - et al.
Less is more: Expectation sharpens representations in the primary visual cortex
Neuron
(2012) - et al.
The distinct modes of vision offered by feedforward and recurrent processing
Trends in Neurosciences
(2000)
Recovery of 3D volume from 2-tone images of novel objects
Cognition
Illusory contours: A window onto the neurophysiology of constructing perception
Trends in Cognitive Sciences
Functional neuroimaging findings on the human perception of illusory contours
Neuroscience and Biobehavioral Reviews
Distinct roles of the cortical layers of area V1 in figure-ground segregation
Current Biology
Always returning: Feedback and sensory processing in visual cortex and thalamus
Trends in Neurosciences
Primary visual cortex activation on the path of apparent motion is mediated by feedback from hMT+/V5
NeuroImage
Expectation (and attention) in visual cognition
Trends in Cognitive Sciences
A neural representation of prior information during perceptual inference
Neuron
Hierarchical organization and functional streams in the visual cortex
Trends in Neurosciences
Varieties of cognitive penetration in visual perception
Consciousness & Cognition
The proactive brain: Memory for predictions
Philosophical Transactions of the Royal Society B: Biological Sciences
Perception and the reach of phenomenal content
Philosophical Quarterly
Physiology of the retina and visual pathway
Ecological cue-validity of “proximity” and of other gestalt factors
The American Journal of Psychology
Communications between cortical areas of the visual system
Whatever next? Predictive brains, situated agents, and the future of cognitive science
Behavioral and Brain Sciences
How the brain learns to see objects and faces in an impoverished context
Nature
Distributed hierarchical processing in the primate cerebral cortex
Cerebral Cortex
There are no top-down influences on perception
Behavioral and Brain Sciences
Perceiving is believing: A Bayesian approach to explaining the positive symptoms of schizophrenia
Nature Reviews Neuroscience
Visual perception and the statistical properties of natural scenes
Annual Review of Psychology
Contour statistics in natural images: Grouping across occlusions
Visual Neuroscience
Implicit social cognition
Psychological Review
The human visual cortex
Annual Review of Neuroscience
Voluntary action and conscious awareness
Nature Neuroscience
A link between visual disambiguation and visual memory
Journal of Neuroscience
Indeterminate organization of the visual system
Science
Recognition alters the spatial pattern of fMRI activation in early retinotopic cortex
Journal of Neurophysiology
The predictive mind
Cortical feedback improves discrimination between figure and background by V1, V2 and V3 neurons
Nature
A new two-alternative forced choice method for the unbiased characterization of perceptual bias and discriminability
Journal of Vision
Availability: A heuristic for judging frequency and probability
Cognitive Psychology
Psychophysics: A practical introduction
Prior expectations evoke stimulus templates in the primary visual cortex
Journal of Cognitive Neuroscience
Cited by (52)
Professional knowledge and task instruction specificity as influencing factors of prospective teachers' professional vision
2022, Teaching and Teacher EducationCitation Excerpt :Furthermore, activated schemata can have a priming effect on identifying relevant aspects and can guide selective attention processes (Johnston & Dark, 1986; cf. noticing). Moreover, (activated) schemata influence how events are perceived, interpreted, and remembered (Gilboa & Marlatte, 2017; Teufel & Nanay, 2017). Early evidence for top-down control was provided by Yarbus (1967).
Pathways of perceptual primacy: ERP evidence for relationships between autism traits and enhanced perceptual functioning
2021, NeuropsychologiaCitation Excerpt :The positive correlation between P1 and N400 effect sizes is consistent with broader theoretical frameworks in which the results of perceptual processing cascade forward to be the subsequent inputs of cognitive mechanisms, thereby influencing cognition. Distinct top-down processing pathways, in which expectations and cognitive biases influence perceptual experiences have also been demonstrated in the literature (e.g., Dijkstra et al., 2017; Teufel and Nanay, 2017), though the data from the current study do not directly speak to these types of information processing loops. This second explanation is also consistent with EPF theory in that neurotypical individuals are more likely to rely on higher-order cognitive processes than autistic individuals; thus, it is plausible that for neurotypical adults, enhanced perceptual discrimination leads to enhanced semantic discrimination.
Word vs. World Knowledge: A developmental shift from bottom-up lexical cues to top-down plausibility
2021, Cognitive PsychologyCitation Excerpt :The bottom-up flow of information—from the retina to the lateral geniculate nucleus, and through a series of cortical regions—results in increasingly complex representations of the visual field, ranging from retinotopic center-surround cells to cortical representations of motion, shape, and object identity (see Maunsell & Newsome, 1987; Lamme & Roelfsema, 2000 for greater detail). This bottom-up processing stream is influenced by expectations at higher-levels via top-down connections, resulting in context effects, priming, and even some visual illusions (e.g. Berry & Schwartz, 2011; Teufel & Nanay, 2017). Given the ubiquity of top-down processing in cognitive systems, it might seem unlikely that children as old as 4–7 years of age would have a global deficit of this kind.
Effect of stimulus dimension on perception and cognition
2021, Acta PsychologicaCitation Excerpt :There are two approaches in explanation of perception-cognition interaction. With a top-down approach, the perceptual systems are shaped and controlled by cognitive system (Manita et al., 2015; Teufel & Nanay, 2017) (Irwin, 2014). In contrast, with the button-up approach, cognitive processing is shaped by perceptual processing as the doorway of information flow (Edelman et al., 2005).
Aesthetic Experience as Interaction
2023, Journal of the American Philosophical AssociationCross-Modal Interaction via Reinforcement Feedback for Audio-Lyrics Retrieval
2024, IEEE/ACM Transactions on Audio Speech and Language Processing