Qualitative differences in the representation of abstract versus concrete words: Evidence from the visual-world paradigm

Abstract

In the present visual-world experiment, participants were presented with visual displays that included a target item that was a semantic associate of an abstract or a concrete word. This manipulation allowed us to test a basic prediction derived from the qualitatively different representational framework that supports the view of different organizational principles for concrete and abstract words in semantic memory. Our results confirm the assumption of a primary organizational principle based on association for abstract words, different from the semantic similarity principle proposed for concrete words, and provide the first piece of evidence in support of this view obtained from healthy participants. The results shed light on the representational structure of abstract and concrete concepts.

Introduction

The representation and processing of concrete versus abstract words has important implications for memory and language theories. Research has shown an advantage for the processing of concrete words (e.g. cross) as compared to abstract words (e.g. faith), referred to as the concreteness effect. Compared to abstract words, concrete words show a recognition benefit in behavioural tasks and eye movement studies (e.g. Bleasdale, 1987, James, 1975, Juhasz and Rayner, 2003), in electrophysiological correlates (e.g. Kounios & Holcomb, 1994), and in brain activation (e.g. Binder, Westbury, McKiernan, Possing, & Medler, 2005).

There have been different proposals to account for this processing advantage of concrete words. Most frameworks assume a quantitative difference, based on the amount of information available when processing abstract or concrete words. For example, it has been proposed that abstract words lack the sensory referents of concrete words (e.g. Paivio, 1986), or that concrete words benefit from the greater availability of related contextual information (e.g. Schwanenflugel & Shoben, 1983). Furthermore, some authors claim that concrete words are supported by a higher number of semantic features (e.g. Plaut and Shallice, 1991, Plaut and Shallice, 1993). However, a qualitative difference in the organization of concrete and abstract words in the mental lexicon has recently been proposed (e.g. Crutch, 2006, Crutch et al., 2006, Crutch and Warrington, 2005, Warrington and Crutch, 2007). In this view, concrete words are primarily organized following a semantic similarity principle, whereas abstract words are mainly organized by their association with other words. According to this view, the primary organizational principle for concrete words is categorical but not associative. In contrast, the representation of abstract words is assumed to rely primarily on semantic association rather than similarity.

This claim has received empirical support from patients with stroke aphasia who show a semantic interference effect (see Crutch & Warrington, 2005). In short, patients showed greater interference for abstract words embedded in an array of words organized by association (e.g. theft in an array that included punishment) as compared to an array created by categorical or synonymic relationships (i.e. semantic similarity; e.g., theft in an array that included burglary). The reverse pattern, however, was found for concrete words: concrete words showed greater interference in arrays defined by semantic similarity than in arrays defined by association.

The qualitatively different representational framework (QDR, for short) assumes that “abstract concepts are represented in an associative neural network whereas concrete concepts have a categorical organization” (Crutch & Warrington, 2005, p. 623). Put differently, abstract words are assumed to be organized mainly by semantic association and concrete words mainly by semantic similarity.¹ Hence, the theory predicts that, once activated, an abstract word would predominantly co-activate associated concepts, whereas a concrete word would predominantly co-activate semantically similar concepts. Accordingly, abstract words are expected to activate their associates faster than concrete words, since this is their primary organizational principle. The present experiment tested this hypothesis using the visual-world paradigm on healthy participants. This study is particularly relevant for two main reasons. First, the confirmation of such an effect within the population of healthy perceivers is important for generalization purposes. Previous evidence has only been obtained from patients suffering from semantic impairments. Considering that the QDR hypothesis is assumed to hold for intact semantic networks as well, a demonstration of its validity in healthy participants would be a vital addition to existing evidence in support of the theory. Second, the QDR framework predicts that differential association effects between abstract and concrete words should appear during on-line word recognition, which requires a paradigm capable of detecting such differences in real-time. The visual-world paradigm is well suited for these purposes.

In this paradigm, an auditory sentence is presented together with a visual scene in which different entities are depicted, while the eye movements of the participant are tracked (e.g. Cooper, 1974, Tanenhaus et al., 1995). The eye movement patterns are affected by some of the properties of the linguistic input. Typically, a critical word in the sentence is related to one of the depicted elements and the proportion of looks to this item provides an index of the strength of the link between the picture and the related auditory input. The variety of relationships manipulated between the auditory input and the depicted items can range from form to syntactic or semantic levels (e.g. Allopenna et al., 1998, Dahan et al., 2001, Huettig and Altmann, 2005, Kamide et al., 2003, Kamide et al., 2003).

In the present study, we will focus on the semantic competition effects shown by Huettig and Altmann (2005). In their experiment, participants were presented with auditory sentences including a critical word (e.g. piano) while looking at a visual display comprising various objects in different quadrants. Interestingly, one of the depicted objects referred to a semantic competitor of the spoken word (trumpet). Participants fixated this object more than any of the other objects, which were unrelated distractors. The authors concluded that “hearing ‘piano’ activated semantic information which overlapped with the semantic information encoded within the mental representation of the concurrent trumpet” (p. B30). In other words, semantically related visual items become active when recognizing a word, and thus capture participants’ visual attention. For this to be possible, visual attention must be guided by the link between specific properties of the visual input and the (auditory) linguistic input, as stated in the so-called linking hypothesis (Tanenhaus, Magnuson, Dahan, & Chambers, 2000). Recently, Altmann and Kamide (2007) developed this proposal further to provide a more detailed account of the dynamics of eye movements in a visual-world experiment. Their conceptual overlap linking hypothesis assumes that the eyes move rapidly towards visual objects whose conceptual representations overlap with those of the objects named in the unfolding linguistic input (see also Dahan and Tanenhaus, 2005, Huettig and Altmann, 2005). In visual-world experiments, the visual input is typically available before the onset of a critical spoken word. Participants’ pre-inspection of the visual scene leads to a pre-activation of conceptual (presumably feature-based) representations of the depicted objects, thereby leaving conceptually enriched episodic traces associated with different object locations in the perceptual experience. Next, with the unfolding of the critical part of speech, these pre-activated representations make contact with the conceptual representations activated by the linguistic input itself – the linguistic input effectively re-activates semantically related episodic traces. Altmann and Kamide (2007) proposed that this leads to a shift in visual attention such that perceivers are more prone to make a saccadic eye movement towards the location of an object that is semantically related to the spoken input. Consequently, the greater the conceptual overlap between a visually presented object and a critical word in the sound stream, the greater the probability of a saccade “back to” the visual object. Huettig, Quinlan, McDonald, and Altmann (2006) provided support for the conceptual overlap linking hypothesis, by showing that semantic proximities in a multi-dimensional space of conceptual representations of visual and auditory stimuli (as derived from Latent Semantic Analysis or contextual similarity measures) are indeed a reliable predictor of eye movements in a visual-world experiment. In this sense, conceptual proximity effects obtained in visual-world experiments can be understood as an analogue to semantic/associative priming effects obtained in word recognition tasks (e.g. doctor activates nurse more than butter).

In the present visual-world experiment, Spanish participants were presented with spoken sentences containing a critical word that could be either concrete or abstract, and with concrete visual scenes that included a critical target object, together with three distractors. In the conditions of primary interest, the critical spoken words were semantic associates of the visual target (e.g. hearing priest and seeing a cross; hearing happiness and seeing a smile). Using the conceptual overlap linking hypothesis (Altmann & Kamide, 2007) and the different organizational principles proposed by the QDR framework (Crutch & Warrington, 2005) as a theoretical basis of our investigations, we aimed at exploring how mental representations activated by concrete versus abstract words link with conceptual representations of associated visual target objects over time. If abstract words are indeed more strongly linked to their relevant associates than concrete words, we should expect more looks towards the associated pictures when hearing an abstract word than when hearing a concrete word. In other words, when hearing an abstract word the spread of activation would primarily flow to associatively related concepts, whereas activation would primarily be spread to semantically similar concepts in the case of hearing a concrete word. Moreover, the visual-world paradigm will enable us to trace the time course of abstract versus concrete word processing.