Horne and Lowe (1996) define naming as a higher order bidirectional relation that incorporates listener and speaker functions such that objects given the
same name enter into a class and become functionally related. Behaviour trained to a subset of this class may generalise to other members without direct training. A key prediction of the naming account is that naming is necessary for the categorisation of formally unrelated objects and events. A series of studies was conducted to tests this prediction by attempting to demonstrate categorisation in the absence of naming behaviours. The behavioural measure of categorisation adopted was categorisation by generalisation; this tested whether behaviour trained to a subset of a potential stimulus class generalised to untrained members of the potential class.
Study 1a investigated whether children between 2.5 and 4 years showed
categorisation of formally unrelated stimuli following only common listener training; four out of four children succeeded on a naming test (tact test) and demonstrated categorisation. Study 1b tested for the extension of classes from three to six member classes; all three participants demonstrated naming and extension of classes. Study 2 investigated whether children of a similar age group showed categorisation following only common speaker training; all four children succeeded on a naming test (listener behaviour test) and demonstrated categorisation. Study 3 further examined these issues using a modification of the design used in Study 1a and 2.
Study 4 (common listener training) replicated Study la with six children under 2.5 years old; two failed a naming test and failed to categorise. Study 5 (common speaker training) likewise replicated Study 2 with three children under 2.5 years old; all three children passed a naming test and categorised.
Studies 1 a through 5 incorporated a naming test prior to testing for
categorisation. Studies 6 and 7 investigated whether the temporal position of the
naming tests affected the incidence of naming behaviour. Study 6 (common listener training) replicated Study I a with four children, except that the naming test occurred after the categorisation tests; two children failed to categorise and both failed to name. Similarly, Study 7 (common speaker training) replicated Study 2 with three children except that the naming test occurred after the categorisation tests; all three categorised and named.
Studies 8a, 8b, 9a, and 9b examined the generalisation of vocal behaviours
following either baseline common listener training (8a & 8b) or common speaker
training (9a & 9b) through the gestural modality; all four children demonstrated
categorisation and named.
Chapter 7 summarises all the studies' findings and discusses their implications for behaviour analytic accounts of categorisation. Taken together, the findings suggest that children who name formally unrelated objects can categorise them, and children who do not name formally unrelated objects cannot categorise
them. These results provide correlational support Horne and Lowe's (1996)
account of naming, and they extend the data on stimulus class formation in infants.