There are fundamental and well-established differences between cognition and perception independent of cognitive penetrability. We will briefly outline properties that distinguish cognition from perception following Halford et al.'s (Reference Halford, Wilson and Phillips2010) contention that relational knowledge is the foundation of higher cognition.
Relational knowledge includes bindings between a symbol and an ordered set of elements. Thus, the relation “elephant larger_than mouse” entails a binding between the relation symbol “larger_than” and slots for a larger and a smaller element (in this example, filled with elephant and mouse, respectively). One of the core properties of higher cognition and not perception is the assignment to slots in a relational structure. Relational knowledge enables structure-based mapping between representations, including analogical reasoning, which is now established as fundamental to reasoning and the acquisition of many concepts (Gentner Reference Gentner2010). Halford et al. (Reference Halford, Wilson, Andrews and Phillips2014) have shown that higher cognition is distinguished by structure-based mapping between representations, enabling inferences.
Symbols, which are fundamental to higher cognition, including both language and reasoning, are core properties of higher cognition. However symbols also require an operating system, which in turn depends on representation of structure. Therefore, to understand “cat eats mouse” we have to assign “cat” to the agent role and ‘mouse” to the patient role. Thus, symbols depend on assignment to slots in a structure, a property that is shared with relational knowledge.
Inferences generated from representations that capture information in premises (Goodwin & Johnson-Laird Reference Goodwin and Johnson-Laird2005) are another core property of higher cognition: For example b larger_than c, a larger_than b can be integrated into a mental model in which a, b, and c are ordered (equivalent to monotonically_larger (a, b, c), permitting the reasoner to make the inference a larger_than c) . Inferences can be generated that are not perceived. Much of reasoning depends on mental models (Khemlani & Johnson-Laird Reference Khemlani and Johnson-Laird2012),
Involvement of working memory, which plays a distinctive role in forming representations, including assignment to slots in a coordinate system (Oberauer Reference Oberauer and Ross2009). The transition from subsymbolic to symbolic cognition depends on assignment of elements to slots in a coordinate system in working memory (Halford et al. Reference Halford, Andrews, Phillips and Wilson2013).
Capacity limits of cognition are different from those of perception, and as humans, we are restricted to representing links between four variables (a quaternary relationship) or fewer in a single cognitive representation (Halford et al. Reference Halford, Baker, McCredden and Bain2005). No such limitation applies to perception.
Other properties of higher cognition have also been identified with relational knowledge (Halford et al. Reference Halford, Wilson, Andrews and Phillips2014). These properties include compositionality, which means preserving components in a compound representation: For example, in “John loves Mary,” “John” retains its identity in the compound representation. Another fundamental property of higher cognition is systematicity, which means that representations are linked by common structure, so the ability to understand the proposition “John loves Mary” is linked to the ability to understand structurally similar propositions such as “Mary loves John” (Fodor & Pylyshyn Reference Fodor and Pylyshyn1988; Phillips & Wilson Reference Phillips, Wilson, Calvo and Symons2014).
Developmental time courses are another point of difference. The development of higher-order cognitive functions exhibits a much slower time course than the development of visual perception. The basic functions of human vision: acuity, contrast sensitivity, stereopsis, and the perception of two-dimensional and three-dimensional shape and size all have reached near-adult levels within the first year (Boothe et al. Reference Boothe, Dobson and Teller1985). More important, even the most methodologically rigorous studies have shown that perceptual organisation following Gestalt principles (e.g., good continuation and good form) is clearly evident in the visual behaviour of 9-month-olds (Quinn & Bhatt Reference Quinn and Bhatt2005; Spelke et al. Reference Spelke, Breinlinger, Jacobson and Phillips1993). Finally, early work claimed that the most “cognitive” of percepts – Kanizsa's illusory contours – is evident in 7-months-olds (Berthenhal et al. Reference Berthenhal, Campos and Haith1980). Though later work has challenged this very young age (Nayar et al. Reference Nayar, Franchak, Adolph and Kiorpes2015), what is not in dispute is that by 7 years of age, the functionality of human visual perception is indistinguishable from that of an adult.
A confused mass of literature has developed as a result of a lack of unifying concepts, and the consequent lack of conceptual coherence, or a paradigm. As a result, we lack tools for interpreting empirical findings. Motivation to reduce everything to a single set of core processes, common to a number of fields including cognitive development, blurs distinctions and means we fail to recognise core properties. The result is that we cannot slice nature at its joints. We can, however, identify the major categories of cognitive processes by focusing on their core properties (Halford et al. Reference Halford, Wilson, Andrews and Phillips2014). Construction of representations in which elements are assigned to slots in a structure, and consequent ability to form structure-consistent mappings between representations, lie at the core of higher cognition. These abstract mappings are neither required nor evident in perception.
We are slicing nature at a joint by acknowledging the distinctiveness of major categories of cognition. To say this is not to deny the significance of processes that are common to many or all cognitions, but it entails acceptance of processes that are distinct to specific functions. Forming structured representations in working memory is a core process that is characteristic of higher cognition. Acceptance of these distinctions leads to a clarity that contributes to the conceptual coherence of the discipline.
There are fundamental and well-established differences between cognition and perception independent of cognitive penetrability. We will briefly outline properties that distinguish cognition from perception following Halford et al.'s (Reference Halford, Wilson and Phillips2010) contention that relational knowledge is the foundation of higher cognition.
Relational knowledge includes bindings between a symbol and an ordered set of elements. Thus, the relation “elephant larger_than mouse” entails a binding between the relation symbol “larger_than” and slots for a larger and a smaller element (in this example, filled with elephant and mouse, respectively). One of the core properties of higher cognition and not perception is the assignment to slots in a relational structure. Relational knowledge enables structure-based mapping between representations, including analogical reasoning, which is now established as fundamental to reasoning and the acquisition of many concepts (Gentner Reference Gentner2010). Halford et al. (Reference Halford, Wilson, Andrews and Phillips2014) have shown that higher cognition is distinguished by structure-based mapping between representations, enabling inferences.
Symbols, which are fundamental to higher cognition, including both language and reasoning, are core properties of higher cognition. However symbols also require an operating system, which in turn depends on representation of structure. Therefore, to understand “cat eats mouse” we have to assign “cat” to the agent role and ‘mouse” to the patient role. Thus, symbols depend on assignment to slots in a structure, a property that is shared with relational knowledge.
Inferences generated from representations that capture information in premises (Goodwin & Johnson-Laird Reference Goodwin and Johnson-Laird2005) are another core property of higher cognition: For example b larger_than c, a larger_than b can be integrated into a mental model in which a, b, and c are ordered (equivalent to monotonically_larger (a, b, c), permitting the reasoner to make the inference a larger_than c) . Inferences can be generated that are not perceived. Much of reasoning depends on mental models (Khemlani & Johnson-Laird Reference Khemlani and Johnson-Laird2012),
Involvement of working memory, which plays a distinctive role in forming representations, including assignment to slots in a coordinate system (Oberauer Reference Oberauer and Ross2009). The transition from subsymbolic to symbolic cognition depends on assignment of elements to slots in a coordinate system in working memory (Halford et al. Reference Halford, Andrews, Phillips and Wilson2013).
Capacity limits of cognition are different from those of perception, and as humans, we are restricted to representing links between four variables (a quaternary relationship) or fewer in a single cognitive representation (Halford et al. Reference Halford, Baker, McCredden and Bain2005). No such limitation applies to perception.
Other properties of higher cognition have also been identified with relational knowledge (Halford et al. Reference Halford, Wilson, Andrews and Phillips2014). These properties include compositionality, which means preserving components in a compound representation: For example, in “John loves Mary,” “John” retains its identity in the compound representation. Another fundamental property of higher cognition is systematicity, which means that representations are linked by common structure, so the ability to understand the proposition “John loves Mary” is linked to the ability to understand structurally similar propositions such as “Mary loves John” (Fodor & Pylyshyn Reference Fodor and Pylyshyn1988; Phillips & Wilson Reference Phillips, Wilson, Calvo and Symons2014).
Developmental time courses are another point of difference. The development of higher-order cognitive functions exhibits a much slower time course than the development of visual perception. The basic functions of human vision: acuity, contrast sensitivity, stereopsis, and the perception of two-dimensional and three-dimensional shape and size all have reached near-adult levels within the first year (Boothe et al. Reference Boothe, Dobson and Teller1985). More important, even the most methodologically rigorous studies have shown that perceptual organisation following Gestalt principles (e.g., good continuation and good form) is clearly evident in the visual behaviour of 9-month-olds (Quinn & Bhatt Reference Quinn and Bhatt2005; Spelke et al. Reference Spelke, Breinlinger, Jacobson and Phillips1993). Finally, early work claimed that the most “cognitive” of percepts – Kanizsa's illusory contours – is evident in 7-months-olds (Berthenhal et al. Reference Berthenhal, Campos and Haith1980). Though later work has challenged this very young age (Nayar et al. Reference Nayar, Franchak, Adolph and Kiorpes2015), what is not in dispute is that by 7 years of age, the functionality of human visual perception is indistinguishable from that of an adult.
A confused mass of literature has developed as a result of a lack of unifying concepts, and the consequent lack of conceptual coherence, or a paradigm. As a result, we lack tools for interpreting empirical findings. Motivation to reduce everything to a single set of core processes, common to a number of fields including cognitive development, blurs distinctions and means we fail to recognise core properties. The result is that we cannot slice nature at its joints. We can, however, identify the major categories of cognitive processes by focusing on their core properties (Halford et al. Reference Halford, Wilson, Andrews and Phillips2014). Construction of representations in which elements are assigned to slots in a structure, and consequent ability to form structure-consistent mappings between representations, lie at the core of higher cognition. These abstract mappings are neither required nor evident in perception.
We are slicing nature at a joint by acknowledging the distinctiveness of major categories of cognition. To say this is not to deny the significance of processes that are common to many or all cognitions, but it entails acceptance of processes that are distinct to specific functions. Forming structured representations in working memory is a core process that is characteristic of higher cognition. Acceptance of these distinctions leads to a clarity that contributes to the conceptual coherence of the discipline.