The Effect of Scene Variation on the Redundant Use of Color in Definite Reference

Abstract

This study investigates to what extent the amount of variation in a visual scene causes speakers to mention the attribute color in their definite target descriptions, focusing on scenes in which this attribute is not needed for identification of the target. The results of our three experiments show that speakers are more likely to redundantly include a color attribute when the scene variation is high as compared with when this variation is low (even if this leads to overspecified descriptions). We argue that these findings are problematic for existing algorithms that aim to automatically generate psychologically realistic target descriptions, such as the Incremental Algorithm, as these algorithms make use of a fixed preference order per domain and do not take visual scene variation into account.

Models and empirical data for the production of referring expressions

“…exclude all other possible entities). By contrast, Koolen, Goudbeek, and Krahmer (2013) found that speakers, when confronted with a blue ball and a blue square (and no objects with a different colour) would still on occasion produce a description like the blue ball (though less frequently than when the square was not blue), suggesting that preference for colour is not completely overruled by discriminatory power even in such simple situations.…”

Developmental Changes in Children’s Processing of Redundant Modifiers in Definite Object Descriptions

“…This way, due to increasing linguistic experience, they learn when to draw an implicature, but also to recognize the situations in which they should not do so. This line of reasoning may in particular be true for absolute modifiers such as color, since these are very frequently used as a result of speaker-internal, saliency-based processes (e.g., Pechmann, 1989; Koolen et al, 2013), without the intention to evoke a specific implicature.…”

“…In those cases, speakers seem to overspecify routinely, for speaker-internal reasons. Previous work in this direction has for example revealed that overspecification is triggered by the presence of visual clutter (Koolen et al, 2016), the incremental nature of speech production (Pechmann, 1989), and the amount of visual variation in the scene (e.g., Clarke et al, 2013; Koolen et al, 2013). In these cases, speakers do not necessarily take the listener perspective into account, but include redundant attributes that are perceptually salient and grab their attention.…”

“…This is what happens in the example where a mother mentions the size of a sweet to warn her child. On the other hand, as noted earlier, redundant attributes can also be used routinely: speakers may, for example, use redundant attributes that are visually salient and grab their attention, such as color (e.g., Pechmann, 1989; Koolen et al, 2013). For children, the question is whether they are able to monitor the speaker’s communicative intentions: at some point in development, they must learn to decide when a redundant attribute is relevant in a given communicative context, and when it is not.…”

Improving Route Directions: The Role of Intersection Type and Visual Clutter for Spatial Reference

“…For example, the more complex or detailed the visual environment, the longer it took participants to start typing their response and resulted in the production of more complex constructions and in more complex eye-tracking patterns (Coco & Keller, 2009). When referring to objects, the amount of visual variation increases the redundant use of attributes (Koolen, Goudbeek, & Krahmer, 2013a). Moreover, in cluttered scenes, referring to a particular object lowers the chance of speakers comparing the object they need to describe to all the other objects (of the same type) present in the scene (Koolen, Krahmer, & Swerts, 2013b).…”