The Effects of Contextual Cues on Visual Categorization

Only available on StudyMode
  • Topic: Reaction time, Presentation of Mary, Object
  • Pages : 13 (4234 words )
  • Download(s) : 17
  • Published : April 4, 2013
Open Document
Text Preview
The Effects of Contextual Cues on Visual Categorization

Ken Sim

Department of Psychology, Department of Engineering
National University of Singapore

2012/2013 Semester 1
The Effects of Contextual Cues on Visual Categorization

Zi Yang Ang, Yan Ting Chan, Ken Sim, Chaoqian Chen, Yu Chen, Mayank Ashok Jain

Department of Psychology, Department of Engineering
National University of Singapore

2012/2013 Semester 1

Abstract
“Do you spot the animal faster than the bird?” Ever since the early studies by Eleanor Roach in the late 70s, much work has been done on categorization. Views on early processing have shifted from “basic level advantage” to “superordinate level advantage”. Recently, Mack and Palmeri claim that superordinate level advantage is only present at short stimulus presentation times. They suggest that at short presentation times, categorization relies more on contextual cues provided by the background, rather than actual object information. We aim to find if the superordinate level advantage at short presentation times can be attributed to contextual cues. We compared human reaction times when categorizing bird pictures as containing either an animal (superordinate level), or bird (basic level) in 3 background types (coherent, incoherent, neutral). Human subjects require an additional 15ms to categorize a bird as a bird (basic level) than as an animal (superordinate). Human subjects also showed the fastest categorization when objects are in neutral backgrounds, followed by coherent and finally, objects in incoherent backgrounds are slowest at being categorized. There is no significant interaction between categorization level and background types showing that the superordinate level advantage might not be due to contextual cues. Introduction

The field is split with differing opinions on perceptual categories. In 1976, Rosch and colleagues proposed a basic level advantage – i.e. we access basic level perceptual categories faster than superordinate level perceptual categories. Murphy and Wisniewski supported this basic level advantage (1989). However, when presented in a scene, there is a decrease basic level advantage while categorizing the object. Building upon that, researchers have found a superordinate level advantage during visual categorization both in isolated contexts (Large et. al., 2004) and in natural scenes (Mace et. al., 2009). Recently, Mack and Palmeri report findings that the superordinate level advantage is only present at short stimulus durations of approximately 25ms (2011). They suggest the possibility that at short stimulus presentation times, there is a greater reliance on coarse visual background information (contextual cues from the picture background) instead of information provided by the object. Soon after, Poncet and colleagues claim that, contrary to Mack and Palmeri’s findings, stimulus presentation duration does not affect the superordinate level speed advantage and at various presentation times (25ms, 250ms or 500ms), reactions times were faster for superordinate level than for basic level categorization (2012). Our group is interested in finding out if the superordinate level advantage can be attributed to contextual cues. Taking note of Poncet and colleague’s comments (that the superordinate level advantage could be affected by the type of stimuli or the type of task), we considered Mack and Palmeri’s results (that superordinate level advantage might be attributed to contextual cues) and we find 3 interpretations for the Poncet et. al. experiments: 1. For both short and long stimulus presentation times, we rely on contextual cues for visual categorization and superordinate level advantage persists for both short and long presentation times. This is supported by Mack and Palmeri’s proposal. 2. For short stimulus presentation times, we rely on contextual cues for visual categorization. However at longer presentation times, we would rely on something else...
tracking img