Similarity and metaphor: A cognitive analytic stance

In several cultures, the sun is the source of ...
Image via Wikipedia

“Any function in the child’s cultural development appears twice, or on two planes. First it appears on the social plane, and then on the psychological plane.” (Vygotsky, 1961: p. 163)


This short essay describes evidence admitting from cognitive developmental studies on the acquisition and understanding of relational similarity in young children and frames the above within salient contemporary debates.

From structural attributes to relational similarity

How is it that the very young can draw comparative inferences in a manner which suggests that they appreciate relational similarity? How do multifactorial processes combine allowing for the ability to map relations by analogical means (e.g., A:B::C:D where short is to tall as thin is to …x)? Much may be owed to the legacy of constructivist influence (e.g., Darwin 1998; Hegel 1969; Helmholtz, 1926; Kant 1999), as many early pioneer developmental theorists (e.g., Erikson, 1950, 1959, 1968, 1969; Inhelder & Piaget, 1958; James, 1982; Piaget, 1970, Piaget, Montangero, & Billeter, 1977; Vygotsky 1961; Winnicott, 1953) held that the presence of categorical knowledge necessitated the inclusion of a structural theory of mind (ToM) to account for differences in terms of staged or graduated process of overcoming. Here an overarching consensus arose between the various schools of thought. That is, at least insofar as quite different structural capacities (stages) were held to exist and demarcate the liminality between the immediate veridical (read concrete) and propositional (read symbolic) knowledge domains attributed to adulthood or childhood respectively.

For Carey (1978; see also Keil & Batterman, 1984; Keil & Caroll, 1980; Keil & Kelly, 1987) young children’s relatively poor ability in tasks requiring the capacity of abstraction (see also Berkeley, 1970; Locke, 1997), from initially specific attributes of measurement (e.g., say, ‘tallness’), were suggestive that more general rule-based categorizations appeared later in the course of the maturational trajectory. In contrast, other researchers found evidence suggesting that young children (3-4yrs) were in possession of a general capacity for abstraction (Baron-Cohen, Leslie, & Frith, 1985; Wimmer & Perner, 1983) or reasoning by inductive inference (i.e., inferences to the best explanation; see also Harman, 1965; Hume, 1748). In other words, with regard to attributes of dimensionality independent of object or context (read abstraction), young children (3-4-5yrs) appeared to some researchers capable of general abstraction moving in the maturational trajectory toward more specific abstract abilities, and to other researchers the opposite view was held. Clearly, therefore, incongruous evidence admitting from studies in defining the processes concerned with characteristic attributional features called out for further research into the underlying putative processes.

Research had amply demonstrated (DeLoache & Brown, 1983, 1984; DeLoache et al., 1985: see also Bower 1982; Bower & Wishart 1972; Diamond 1991, 2000) that very young children demonstrate excellent levels of competence in tasks requiring spatial mapping skills in order to locate occluded objects when compared to infants (3-4mths) poor differential looking times. For Judy DeLoache (1987) a reasonable extension of previous findings appeared not just limited by the extent to which spatial mapping skills (i.e., a furnished room and a scale model of the room) could be demonstrated in very young children (N=32, 16×30-32mths & 16×36-39mths). Rather, DeLoache (1987) proposed that her studies provide an important clue compatible with the suggestion that object knowledge and symbolic representation (symbolization) of the object occurred in coincidence earlier than researchers had previously thought (e.g., Piaget 1970). However, DeLoache (1987: p. 1556) appears to pass over the importance in her results owed to the role her own language played on the reasoning of the young children during an “extensive orientation phase”. In addition, the suggestion might be made that the role of the experimenter’s language with the child appears crucial in the light cast by having “explicitly described and demonstrated the correspondence between the two toys to be occluded, between the room and the model, and between the individual items of furniture (the hiding places) within the two spaces.” In other words, one might reasonably suppose that verbal instruction might easily have influenced the outcomes admitting from this study.

That duly noted, for Halford’s (1987, 1992, 1993) structure-mapping class inclusion tasks on the basis of relational reasoning, that is, reasoning by abstract analogy, results (i.e., three-coloured pegs to three-coloured rods of different lengths) seemed compatible with suggestions in favour (e.g., Rattermann, 1991; Rattermann, Gentner, & DeLoache, 1989, 1994) of perceptual categorization inferred by means of a domain general increase in analogical capacity (i.e., domain general account). Contradistinctively, Usha Goswami (1993; see also Goswami & Brown, 1989) proposed that perceptual categorization knowledge might also be adequately accounted for through the acquisition of domain specific knowledge at various intervals (i.e., domain specific account). To point up her case, Goswami (1995) examined the analogical reasoning (transitive relational-mapping) ability of young children (3-4yrs) based on the perceptual categorization of a ternary relation (i.e., pairs of relations linking three elements) along a single dimensional plane (i.e., size of cups using the analogy of ‘daddy’, ‘mummy’ and ‘baby’). Here Goswami’s (1995) findings suggest an interpretation which places emphasis not on the structural capacity for recognition of relational similarity per se (i.e., domain general thesis) but rather emphasises the prior knowledge of a representation for those relations (i.e., say, the analogy of the three bears). Goswami (1995) goes on to suggest that these findings remain consistent whether the objects being mapped are similar or dissimilar (e.g., Gentner & Rattermann, 1991), or based upon binary or ternary relations (e.g., Halford, 1987, 1993).

Analogical Development

For Humbert Maturana & Francisco Varela (1980) a metonymic process (i.e., a metaphorical or analogical connection linking more than one object to a similar representation) relies upon the competence of the individual and the rational-linguistic abilities to compare between two registers, that is, two phenomenological domains (Bowdle & Gentner, 2005; see also Vygotsky, 1961; Winnicott, 1953): one concrete and one symbolic domain of representational knowledge. Comparisons between either of these two domains of knowledge cannot simply be limited to thought per se as if in isolation from language; thus, the role of language in metaphorical and analogical reasoning alike suggests itself as a necessary component in the development of the mature comparative capacity for relational similarity. It is perhaps the competence requirement of metonymy which perhaps most clearly suggests for some researchers (e.g., Bowdle & Gentner, 2005; Maturana & Varela, 1980) that, whether metaphorical or analogical, cross-domain mapping will become spontaneously instantiated or conventional.

To Jeffrey Loewenstein & Dedre Gentner’s (2001) thought their findings suggest that the process of comparison of surface features within and between domains (see also Gentner, 1983; Kotovsky & Gentner, 1996; Rattermann & Gentner, 1998) might be one end of a spectrum of factors which facilitates the analogical spatial mapping abilities of children. On the other side of this spectrum, so to speak, Meints, Plunkett, Harris, & Dimmock (2002) tested spatial prepositions using preferential looking tasks. Here infants were studied to see whether they might comprehend the use of simple prepositions (i.e., ‘in’, ‘on’, and ‘under’) applied in a variety of typical and atypical situations. Meints et al. (2002) results confirm the predicted outcome that these simple prepositions (i.e., ‘on’ and ‘under’) are two of the first to be understood (measured based on looking time). Moreover, Meints et al. (2002) sample population (15-18-24mth) of infants appears to suggest that the longer looking times recorded in the 15-18mth age range would be entirely consistent with typical image naming as the association between the newly acquired word and spatial relationship of the object was not yet fully cemented. Further, Meints et al. (2002) propose that their findings offer a buttress to the suggestion that young children utilize not dissimilar learning strategies for mapping object naming and spatial preposition acquisition (see also Hupp & Haaf, 2004). Never the less, the ontogenesis of the proposed suggestion for an economical learning strategy (Meints et al. 2002) remains largely unsatisfactory; the point of origination for such a learning strategy might equally well be attributed to parental input as it might be suited to children’s prelinguistic conceptual structures, both or indeed neither.

For Robert Leech, Denis Mareschal, & Richard Cooper (2003) analogical reasoning per se acts as an umbrella term for a variety of core cognitive and developmental processes. To Leech et al. (2003: p. 710) analogical reasoning does not require specialized processes to develop; a simple memory store and a gradual accretion of knowledge may be sufficient to produce instantial analogies: “No special mechanisms are required to deal with analogy.” Successful connectionist modelling (i.e., associationistic; see also McClellend & Rummelhart, 1986) of the development of analogical reasoning ought, instead, to directly address certain key phenomena: the shift from judging similarity from attributes to relations (after Gentner, 1988; Gentner & Rattermann, 1998; Gentner & Toupin, 1986); the robust role of relational epistemology underpinning the development of analogical inferences by induction (after Goswami & Brown, 1989; Gentner & Rattermann, 1998); the domain specific grounding for the capacity over development toward domain specific instantial analogies (after Goswami & Brown, 1989); spontaneous analogical ability without directed tutelage (after Paen & Wilkening, 1997; Ingaki & Hatano, 1987); and, a greater focus on the temporal role in knowledge acquisition (i.e., epistemology and time). The model proposed by Leech et al. (2003) suggests how a complex cognitive reasoning skill might emerge from a simple causal system or apparatus (see also Goswami, 2001). None the less, although it is clear that learning from transformational errors does represent a valid alternative to a structural, domain general increase in cognitive capacity sufficient to inaugurate analogical reasoning, the connectionist account cannot rid itself of a foundational overextension. To use an analogy, this possible overextension affirms that, say, a monkey with a typewriter could over many epochs of time produce works akin to a Proust or a Shakespeare. It is this core connectionist overextension (i.e., or the inherent constraint; Goswami & Pauen, 2005: p. 123n) which, perhaps, renders absurd the richness of adult development seen exclusively through the conduit offered by the early connectionist learning standpoint.

That having been said, Leech, Mareschal, & Cooper (2008) revisit developmental thought with a more subtle and tempered suggestion that, while essentially a simple apparatus (i.e., similar to those found in childhood), analogical reasoning might function to establish complex patterns of relational priming (i.e., similar to those found in adulthood; see Tulving & Schacter, 1990). For Leech et al. (2008) the revised key phenomena most in urgent need of consideration are twofold: to situate development as the kernel for any account of cognition; and, the necessity for such models to demonstrate development across the full maturational trajectory; that is, from simple to complex analogical systems. Moreover, Leech et al. (2008: p. 378) are expressly mindful of the inadequacies of their current relational priming model, and anticipate a considerable temporal aporia before “a very different model is arrived at” capable of such complex life-cycle modelling. Such a paradigm shift in connectionist modelling is difficult to imagine at this time; though, models of this kind would of course be required to demonstrate understanding at the levels of the semantic ‘meaning’ of its component terms and also the syntactic outcomes from countless iterations of possible analogies.

However, the assertion of functional inadequacy constraining current models perhaps renders research in this field an effective extension of functional capability rather than merely marking an overestimation of current capabilities. In this way such models of relational priming surely act to buttress the suggestion that sustains the investigation of relational priming prototypes, and even points up the significance of development in psycho-biological connectionist modelling architectures per se. In this light, connectionist modelling promises a telos for developmental prototyping (McClellend & Rummelhart, 1986; Gazzaniga, Ivry, & Mangun, 1998). Notwithstanding, the potential utility of relational priming remain open for a great deal of further research.

Reflections in conclusion

In the essay above the reportage has aimed to highlight a small part of the degree to which analogical reasoning tasks clarify evidence admitting from the relational similarity paradigm. After some general initial historical observations on stage theories and graduated overcoming theories, the second paragraph centres on the shift in emphasis from the structural attributes (Carey, 1978; Keil & Batterman, 1984; Keil & Caroll, 1980; Keil & Kelly, 1987) to analogy by relational similarity (Baron-Cohen, Leslie, & Frith, 1985; DeLoache & Brown, 1983, 1984; DeLoache et al., 1985; Wimmer & Perner, 1983). The third and fourth paragraphs report on some studies concerned with competence in spatial mapping and the perceptual categorization by mapping via domain general (Halford, 1987, 1993; Rattermann, 1991; Rattermann, Gentner, & DeLoache, 1989, 1994) and domain specific accounts (Goswami, 1993, 1995; Goswami & Brown, 1989). Evidence in the fifth paragraph is considered from the standpoint of analogical development (i.e., analogy, metaphor and metonymy) as constrained by the performance of a specific individual situated within a wider register of general competence (Bowdle & Gentner, 2005; Maturana & Varela, 1980). Data is outlined in the sixth paragraph from within and between domains looking at analogy using both rational thought and language prepositions (Jeffrey Loewenstein & Dedre Gentner’s (2001) thought their findings suggest that the process of comparison of surface features within and between domains (see also Gentner, 1983; Kotovsky & Gentner, 1996; Meints, Plunkett, Harris, & Dimmock, 2002; Rattermann & Gentner, 1998). In the seventh paragraph a short account of an initial outline for a connectionist theory and model of analogical reasoning is considered (Leech, Mareschal, & Cooper, 2003). The eighth paragraph introduces a revised and more nuanced connectionist model proposal; that is, that simple analogical reasoning apparatus might function to establish rich, complex patterns of relational priming (Leech, Mareschal, & Cooper, 2008). By way of conclusion the ninth paragraph closes by addressing the promise (McClellend & Rummelhart, 1986), and potential hazards (Gazzaniga, Ivry, & Mangun, 1998) facing developmental theorists and the utility of current connectionist modelling architectures.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s