Generalisation of Word-Picture Relations in Children with Autism and Typically Developing Children
Due to profound linguistic impairments, many low-functioning children with autism (CWA) are taught to communicate using pictorial symbols. Previous research has shown that CWA are significantly more likely to generalise labels to objects depicted in iconic colour pictures than non-colour pictures. However, it cannot be assumed that CWA privilege the same cues as typically developing children (TDC) when generalising words from colour pictures. From 2 years, TDC selectively generalise labels based on sameness of shape – the fundamental constraint underlying word-picture-object relations. By contrast, CWA often have difficulty categorising around the dimension of shape, and many fail to develop a ‘shape bias’ word learning heuristic.
Our objective was to identify whether CWA generalise labels from colour pictures based on similarity of shape, colour, or both shape and colour. We predicted that the absence of a shape bias may cause some CWA to over-extend labels to objects that match depicted referents on colour, but not shape.
Seventeen CWA (M age = 9.7 years) were matched to 17 TDC (M age = 3.5 years) on receptive language (CWA: M = 3.5 years; TDC: M = 3.5 years). In each of 4 test trials, children were taught a novel word paired with a colour photograph of an unfamiliar target object, and then sorted a series of items into two containers according to whether or not they were a referent of the newly-learned label. Sorted items included 3-D target objects, differently-coloured variants of target objects, novel unfamiliar objects that were colour-matched to target objects, familiar objects, and pictures of each type of object.
CWA almost always extended labels to items that matched colour photographs on shape and colour (target picture: 100%; target object: 97.5%), but also frequently generalised to items that matched only on shape (shape-match picture: 72.5%; shape-match object: 69%) or only on colour (colour-match picture: 69%; colour-match object: 61.8%). Two-thirds of CWA extended labels to shape-matched items and colour-matched items on 2-4 trials, suggesting their generalisation of word-picture relations was directed by both shape and colour. By contrast, TDC invariably generalised labels on the basis of shape (target picture: 100%; target object: 98.5%; shape-match picture: 100%; shape-match object: 98.5%), but rarely extended to items matching only on colour (colour-match picture: 11.8%; colour-match object: 7.3%). Neither group generalised words to familiar items.
Low-functioning CWA understand that words paired with colour photographs can be extended to independently existing referents, however, they often generalise based on incorrect dimensions (i.e. colour). This atypical pattern of responding indicates a crucial misunderstanding of the rules that govern referential word-picture-object relations. Deficits in category formation may prevent CWA from privileging shape as a basis for generalising word-picture relations, meaning they are just as likely to extend words to items that match depicted objects on superficial perceptual details, such as colour. These findings have important implications for the design and delivery of picture-based communication interventions.