The Effects of Feature‐Label‐Order and Their Implications for Symbolic Learning

Abstract

Symbols enable people to organize and communicate about the world. However, the ways in which symbolic knowledge is learned and then represented in the mind are poorly understood. We present a formal analysis of symbolic learning-in particular, word learning-in terms of prediction and cue competition, and we consider two possible ways in which symbols might be learned: by learning to predict a label from the features of objects and events in the world, and by learning to predict features from a label. This analysis predicts significant differences in symbolic learning depending on the sequencing of objects and labels. We report a computational simulation and two human experiments that confirm these differences, revealing the existence of Feature-Label-Ordering effects in learning. Discrimination learning is facilitated when objects predict labels, but not when labels predict objects. Our results and analysis suggest that the semantic categories people use to understand and communicate about the world can only be learned if labels are predicted from objects. We discuss the implications of this for our understanding of the nature of language and symbolic thought, and in particular, for theories of reference.Keywords: Language; Learning; Representation; Concepts; Computational modeling; Prediction Symbolic thought and symbolic communication are defining human characteristics. Yet despite the benefits symbols bring in allowing us to organize, communicate about, manipulate, and master the world, our understanding of symbols and symbolic knowledge is poor. Centuries of pondering the nature of symbolic representation, in terms of concepts and categories and words and their meanings, has yielded more puzzles than answers (Murphy, 2002; Wittgenstein, 1953). Our impoverished understanding of symbolic learning, and especially how words and their meanings are learned, represented, and used, contrasts starkly with the progress made in other areas, where computational models of learning processes Correspondence should be sent to Michael Ramscar,

A learning perspective on individual differences in skilled reading: Exploring and exploiting orthographic and semantic discrimination cues.

“…It also supports work of Ellis (2006a;2006b;, as well as work of Ramscar and associates (Ramscar et al, 2010;. These authors have argued that the Rescorla-Wagner model (1972), implemented in our Naive Discrimination Learning framework, does not necessarily respect (fixed) linguistic categories while discriminating form-meaning relationships important for understanding the meaning of an utterance.…”

“…Language appears as a complex information channel (Shannon, 1948), continuously evolving to facilitate communication (more about its complex adaptive nature in Beckner et al, 2009;Ramscar, 2010;Ramsar & Baayen, 2013). Moreover, it unfolds on many levels of granularity, from orthographic and phonological elements, over words and word chunks to sentences and beyond.…”

“…Ramscar, Yarlett, Dye, Denny and Thorpe (2010) have demonstrated that symbolic learning, and word learning in particular, consists in probing which of a sceneÕs semantic features are the best predictors of its appropriate phonological label. Through computational simulation and behavioural experimentation, they showed that children extract a range of semantic information from a scene (e.g.…”

Implicative organization facilitates morphological learning

“…Second, the task involved inferring relations among actual referents, rather than purely among abstract linguistic labels. These referents were presented prior to their labels, and this presentation style is known to enhance association learning (Ramscar et al 2010). Furthermore, participants were required to actively produce a free response to every item, from the very start of the task.…”

The cave of shadows: Addressing the human factor with generalized additive mixed models

“…As shown by Marsolek (2008), the association strengths between visual features and object names are subject to continuous updating. Ramscar et al (2010) and Arnon and Ramscar (2012) documented the consequences of within-experiment learning in the domain of language. Kleinschmidt and Jaeger (2015) report and model continuous updating in auditory processing in the context of speakerlistener adaptation.…”