Occluding the face diminishes the conceptual accessibility of an animate agent

Abstract

The language that people use to describe events reflects their perspective on the event. This linguistic encoding is influenced by conceptual accessibility, particularly whether individuals in the event are animate or agentive--animates are more likely than inanimates to appear as Subject of a sentence, and agents are more likely than patients to appear as Subject. We tested whether perceptual aspects of a scene can override these two conceptual biases when they are aligned: whether a visually prominent inanimate patient will be selected as Subject when pitted against a visually backgrounded animate agent. We manipulated visual prominence by contrasting scenes in which the face/torso/hand of the agent were visible vs. scenes in which only the hand was visible. Events with only a hand were more often associated with passive descriptions, in both production and comprehension tasks. These results highlight the power of visual prominence to guide how people conceptualize events.

Keywords: animacy; event representation; passive voice; sentence production.

Figure 1.

Examples of Body-Agent (A), Hand-Agent (B), and No-Agent (C) events in Experiment 1. In A and B a woman is tipping over a book; in C a book is falling over.

Figure 2.

Proportion of responses coded as active, passive or intransitive in the Body-Agent, Hand-Agent and No Agent conditions, in Experiments 1A-1C (1A: between subjects; 1B: within subjects randomized; 1C: within subjects blocked). Error bars show 95% confidence intervals.

Figure 3.

Examples of Body-Agent (A) and Hand-Agent (B) events from Experiment 2. In the pictured event the woman is shutting a box.

Figure 4.

Proportion of responses coded as active, passive or intransitive in the Body-Agent and Hand-Agent conditions in Experiment 2. Error bars show 95% confidence intervals.

Figure 5.

Proportion of Body-Agent vs. Hand-Agent trials where the Body-Agent video was chosen, for each sentence type in Experiment 3A and 3B. Error bars show 95% confidence intervals.

Figure 6.

Average animacy rating on a 1 – 7 point scale for each of the video types in Experiment 4. Error bars show 95% confidence intervals.