Evidence from numerous studies using the visual world paradigm has revealed both that spoken language can rapidly guide attention in a related visual scene and that scene information can immediately influence comprehension processes. These findings motivated the coordinated interplay account (Knoeferle & Crocker, 2006) of situated comprehension, which claims that utterance-mediated attention crucially underlies this closely coordinated interaction of language and scene processing. We present a recurrent sigma-pi neural network that models the rapid use of scene information, exploiting an utterance-mediated attentional mechanism that directly instantiates the CIA. The model is shown to achieve high levels of performance (both with and without scene contexts), while also exhibiting hallmark behaviors of situated comprehension, such as incremental processing, anticipation of appropriate role fillers, as well as the immediate use, and priority, of depicted event information through the coordinated use of utterance-mediated attention to the scene.

Empirical evidence demonstrating that sentence meaning is rapidly reconciled with the visual environment has been broadly construed as supporting the seamless interaction of visual and linguistic representations during situated comprehension. Based on recent behavioral and neuroscientific findings, however, we argue for the more deeply rooted coordination of the mechanisms underlying visual and linguistic processing, and for jointly considering the behavioral and neural correlates of scene-sentence reconciliation during situated comprehension. The Coordinated Interplay Account (CIA; Knoeferle & Crocker, 2007) asserts that incremental linguistic interpretation actively directs attention in the visual environment, thereby increasing the salience of attended scene information for comprehension. We review behavioral and neuroscientific findings in support of the CIA's three processing stages: (i) incremental sentence interpretation, (ii) language-mediated visual attention, and (iii) the on-line influence of non-linguistic visual context. We then describe a recently developed connectionist model which both embodies the central CIA proposals and has been successfully applied in modeling a range of behavioral findings from the visual world paradigm (Mayberry, Crocker, & Knoeferle, in press). Results from a new simulation suggest the model also correlates with event-related brain potentials elicited by the immediate use of visual context for linguistic disambiguation (Knoeferle, Habets, Crocker, & Münte, 2008). Finally, we argue that the mechanisms underlying interpretation, visual attention, and scene apprehension are not only in close temporal synchronization, but have co-adapted to optimize real-time visual grounding of situated spoken language, thus facilitating the association of linguistic, visual and motor representations that emerge during the course of our embodied linguistic experience in the world.

No abstract

Abstract. Recent "visual worlds" studies, wherein researchers study language in context by monitoring eye-movements in a visual scene during sentence processing, have revealed much about the interaction of diverse information sources and the time course of their influence on comprehension. In this study, five experiments that trade off scene context with a variety of linguistic factors are modelled with a Simple Recurrent Network modified to integrate a scene representation with the standard incremental input of a sentence. The results show that the model captures the qualitative behavior observed during the experiments, while retaining the ability to develop the correct interpretation in the absence of visual input.

No abstract