Rapid natural scene categorization in the near absence of attention

Abstract

What can we see when we do not pay attention? It is well known that we can be "blind" even to major aspects of natural scenes when we attend elsewhere. The only tasks that do not need attention appear to be carried out in the early stages of the visual system. Contrary to this common belief, we report that subjects can rapidly detect animals or vehicles in briefly presented novel natural scenes while simultaneously performing another attentionally demanding task. By comparison, they are unable to discriminate large T's from L's, or bisected two-color disks from their mirror images under the same conditions. We conclude that some visual tasks associated with "high-level" cortical areas may proceed in the near absence of attention.

Figure 1

Experimental protocol. (a) Schematic illustration of one trial. After a fixation cross presented at the center of the visual field, an attentionally demanding letter discrimination task is presented centrally. The central stimulus (combination of five T's and L's) is then replaced by a perceptual mask (five F's) after a time interval commonly called the stimulus onset asynchrony (SOA, ranging from 133 to 240 ms for different subjects). Subjects are instructed to respond whether all five letters are the same or one of them is different. In the peripheral natural scene categorization task, an image is presented peripherally for 27 ms at a random location, 53 ms after the onset of the central stimulus. The peripheral stimulus is followed (after peripheral SOA) by a perceptual mask. The peripheral SOA varies individually for each subject, ranging from 53 to 80 ms. The peripheral mask always appears before the central stimulus is replaced by its own mask. Subjects make a speeded response to the presence of animals. Under the dual task condition, subjects are required to perform both tasks concurrently. (b) Sample images of the stimulus database. The pictures are complex color scenes taken from a large commercially available CD-ROM library allowing access to several thousand stimuli. The animal category images include pictures of mammals, birds, fish, insects, and reptiles. In a separate experiment (Fig. 3 b and c), an additional target category is used—vehicles. The vehicle category images include pictures of cars, trucks, trains, airplanes, ships, and hot-air balloons. There is also a very wide range of distractor images, which includes natural landscapes, city scenes, food, fruits, plants, houses, and artificial objects.

Figure 2

Main results. Individual subject's results for dual vs. single task performance (five subjects). The horizontal axis represents performance of the central task (attentionally demanding letter discrimination task). The vertical axis represents performance of the peripheral task (natural scene categorization). Each filled circle represents the performance of one block (96 trials per block) under the dual task condition. All images used for testing were novel for the subjects. Each open circle represents the average performance under the single task condition. For each subject, performances of the letter discrimination task do not differ significantly (t test, P > 0.05) under the single and dual task conditions, suggesting that attention was fully allocated to the center in the dual task condition. Furthermore, the performances of the natural scene categorization task do not differ significantly (t test, P > 0.05) either under the single and dual task conditions, suggesting that the task may be performed while attention is engaged elsewhere.

Figure 3

(a) Summary of categorization of masked animal images. This panel corresponds to the normalized average performance of the main experiment (Fig. 2). Each open circle is the average value of one subject's dual task performance, normalized according to his/her own single task performance: a linear scaling transforms the average single task performance into 100%, leaving chance at 50%. Performance under the dual task condition that is higher than the corresponding performance under the single task condition would result in a normalized performance higher than 100%. Error bars reflect the standard error of the means. (b and c) Categorization of natural and artificial objects. The same five subjects performed the following two categorization tasks in alternating blocks. (b) Categorization of masked animal images among vehicles and other distractors. Distractors for this task include 50% vehicle scenes and 50% non-animal/non-vehicle scenes, randomly drawn from the same database described in Fig. 1b. Task performance for each of the five subjects is comparable under dual task and single task conditions (t test, P > 0.05). This panel presents a summary of normalized average performance of each subject as detailed in a. (c) Categorization of masked vehicle images. Subjects are instructed to perform the natural scene categorization task by using vehicles as targets (including cars, trucks, trains, airplanes, ships, and hot-air balloons). Distractors for this task include 50% animal scenes and 50% non-animal/non-vehicle scenes. The panel illustrates normalized dual task performances of the five subjects. For each subject, task performance is comparable under dual task and single task conditions (P > 0.05). These experiments provide evidence that artificial and natural target categories can be detected in the near absence of attention. (d and e) Control experiments. (d) Peripheral letter discrimination task. Five subjects are instructed to discriminate between the letters T and L presented in the periphery. The letter, randomly rotated, is masked by the letter F after the peripheral SOA (ranging from 53 to 160 ms). For each subject, this peripheral letter discrimination task cannot be performed above chance in the absence of attention (paired t test, P > 0.05). This panel presents a summary of normalized average performance of each subject. (e) Peripheral color pattern discrimination task. Five subjects are instructed to discriminate a red/green color disk from a green/red color disk. The stimulus is masked after the peripheral SOA (ranging from 66 to 106 ms). For each subject this peripheral color pattern discrimination task cannot be performed above chance in the absence of attention (paired t test, P > 0.05). The results from these control experiments demonstrate that our central discrimination task effectively withdraws attention away from the peripheral task (1). This panel presents a summary of normalized average performance of each subject.