Visual determinants of preferred adjective order

Abstract

In referential communication, speakers refer to a target object among a set of context objects. The NPs they produce are characterized by a canonical order of prenominal adjectives: The dimensions that are easiest to detect (e.g., absolute dimensions) are commonly placed closer to the noun than other dimensions (e.g., relative dimensions). This stands in stark contrast to the assumption that language production is an incremental process. According to this incremental-procedural view, the dimensions that are easiest to detect should be named first. In the present paper, an alternative account of the canonical order effect is presented, suggesting that the prenominal adjective ordering rules are a result of the perceptual analysis processes underlying the evaluation of distinctive target features. Analyses of speakers’ eye movements during referential communication (Experiment 1) and analyses of utterance formats produced under time pressure (Experiment 2) provide evidence for the suggested perceptual classification account.

Acknowledgments

Experiment 1 was conducted when the author was a member of the graduate programme “Task Oriented Communication” at the University of Bielefeld, Germany, and was supported by the Deutsche Forschungsgemeinschaft (DFG; GRK 256/2). It is part of a doctoral dissertation submitted by the author to the Department of Linguistics at the University of Bielefeld in December 2001. Experiment 2 was conducted when the author was funded by a postdoctoral scholarship from the German Academic Exchange Service (DAAD; D/02/00789). Many thanks to Antje Meyer for providing the lab facilities required for this experiment. I wish to thank Antje Meyer, Ulrich Schade, and Hans-Jürgen Eikmeyer for discussions on the presented findings.

Notes

¹Danks and Schwenk (1972; see also Danks & Glucksberg, 1971) formulated a “pragmatic communication rule” to account for ordering preferences in prenominal adjective order. They claimed that ordering effects should follow a principle of relative relevance of the adjectival predications and thus predicted high frequencies of inverted adjective orders in specific situational contexts. Although they present empirical findings that seem to verify these predictions, Richards (1975) reanalysed these findings and showed that there is no conclusive evidence for the pragmatic communication rule (Richards, 1975, 1977).

²As the focus of the experiment lay on the relative order of colour and size adjectives, object class differences were left out in order to keep the number of objects within the display as small as possible for a better quality of the eye movement measurements.

³The relevant sections of the written instructions were held as similar as possible in the two instruction groups by adding instructions on naming the object minimally (in the minimal instruction group; printed in italics) without altering the original neutral instruction: “When the objects appear on the screen, please name the target object as quickly and as accurately as possible and as minimally as possible. You should name it in such a way that a listener who does not know the target object would be able to recognize it on the basis of your description and in such a way that your description does not include any features that are redundant in terms of the unique identification of the target object.”

⁴The rate of size overspecifications appears to be quite high compared to other experiments (see, e.g., Eikmeyer & Ahlsén, 1998). This may be due to the fact that with a free utterance format, participants preferably use full specifications to be exact and therefore include both colour and size, whenever possible. As outlined in the methods section, irrelevant variations of size were present in all trials of the type colour-only, so in this context, overspecifications of size were comparably rare, occurring in only one out of four trials.