Language is activated by visual input regardless of memory demands or capacity

Abstract

In the present study, we provide compelling evidence that viewing objects automatically activates linguistic labels and that this activation is not due to task-specific memory demands. In two experiments, eye-movements of English speakers were tracked while they identified a visual target among an array of four images, including a phonological competitor (e.g., flower-flag). Experiment 1 manipulated the capacity to subvocally rehearse the target label by imposing linguistic, spatial, or no working memory load. Experiment 2 manipulated the need to encode target objects by presenting target images either before or concurrently with the search display. While the timing and magnitude of competitor activation varied across conditions, we observed consistent evidence of language activation regardless of the capacity or need to maintain object labels in memory. We propose that language activation is automatic and not contingent upon working memory capacity or demands, and conclude that objects' labels influence visual search.

Keywords: Cognitive load; Language activation; Phonological competition; Visual search; Visual world paradigm; Working memory.

Figure 1.

Sample trial structure for the No-Load (a), Linguistic-Load (b), and Spatial-Load (c) conditions. The target (e.g., flower) was present in the search display along with a phonological competitor (e.g., flag) and a control and filler item (e.g., knife, cat) which did not overlap phonologically. Participants were instructed to click on the target object as quickly as possible.

Figure 2.

Visual fixation peak shapes. Significant effects of Competition were observed on the linear, cubic, and quartic terms in all three conditions, the combination of which reflect differences in the steepness of the curves for competitors (black) vs. controls (gray). In particular, the rise and fall of fixations around the peak were steeper for competitors than controls in the No-Load (solid lines) and Linguistic-Load conditions (dashed lines). In contrast, fixations to competitors in the Spatial-Load condition (dotted lines) were more gradual and sustained relative to controls. There were additionally significant effects of Competition on the intercept for the No-Load (solid lines) and Spatial-Load (dotted lines) conditions, indicating that overall, competitors were fixated more than controls. Lines represent best fit logistic regression model estimates. Timecourses were re-centered for each subject with 0 ms corresponding to peak activation for each condition.

Figure 3.

Eye-tracking latency across load conditions. Peak latencies revealed comparable rates of peak activation for phonological competitors (black) and controls (gray) in No-Load and Linguistic-Load conditions. In the Spatial-Load condition, competitors peaked later than controls due to more sustained activation of the competitor. Curves represent observed data, dots represent peak latency, and horizontal lines represent standard error.

Figure 4.

Left: Peak shapes revealed reduced target activation in the Spatial-Load condition (dotted line) relative to No-Load (solid line), particularly at and following peak onset (0 ms). Overall fixations to the target were greater under Linguistic-Load (dashed line) than No-Load (solid line). Lines represent best fit logistic regression models. Timecourses were re-centered for each subject with 0 ms corresponding to peak activation for each condition. Right: Peak latencies revealed that target fixations peaked later under Spatial-Load (dotted lines, square), and to a lesser extent under Linguistic-Load (dashed lines, triangle) relative to No-Load (solid lines, circle). Curves represent observed data, dots represent peak latency, and horizontal lines represent standard error.

Figure 5.

Sample trial structure for the Long (a), Short (b), and No-Delay (c) conditions. The target (e.g., flower) was present in the search display along with a phonological competitor (e.g., flag) and a control and filler item (e.g., knife, cat) which did not overlap phonologically. Participants were instructed to click on the target object as quickly as possible.

Figure 6.

Visual fixation peak shapes. Significant effects of competition were found for the intercept in the No-Delay (dotted lines) and Short-Delay conditions (dashed lines), indicating that competitors (black) were fixated more than controls (gray). When the search target was presented at the same time as the visual display (No-Delay, dotted lines), fixation peaks were sustained longer (effects of competition on the linear, quadratic, and cubic terms), whereas advance preview of the target resulted in sharper peaks (effects of competition on the quadratic term for both Short- and Long-Delay). Longer delays between presentation of the search target and the display (solid lines) resulted in greater fixations than a short delay (dashed lines; main effect of Delay-Length on the intercept). Lines represent best fit logistic regression model estimates. Timecourses were re-centered for each subject with 0 ms corresponding to peak activation for each condition.

Figure 7.

Eye-tracking latency across delay times. Peak latencies revealed earlier activation of phonological competitors (black) than controls (gray) in Long-Delay (solid lines, circles) and Short-Delay (dashed lines, triangles) conditions. Competitor and control peak latencies did not differ in the No-Delay condition (dotted lines, squares). Curves represent observed data, dots represent peak latency, and horizontal lines represent standard error.

Figure 8.

Left: Peak shapes revealed reduced target activation in the Short-Delay condition (dashed line) relative to Long-Delay condition (solid line; effect of Delay-Length on the intercept). Significant effects of Delay-Any and Delay-Length on the linear, quadratic, cubic, and quartic terms indicate that the steepness of the fixation curves differed between No-Delay and the two delay conditions, as well as between Short- and Long-Delay in the center and tail ends of the window. Visual inspection suggests that target activation was faster preceding peak onset (0 ms) in the Long-Delay condition (solid line) compared to the Short- (dashed line) and No-Delay (dotted line) conditions. Target activation following peak-onset was more sustained in the No-Delay condition relative to the two delay conditions, and activation was more sustained in the Long-Delay condition relative to the Short-Delay condition. Lines represent best fit logistic regression models. Timecourses were re-centered for each subject with 0 ms corresponding to peak activation for each condition. Right: Peak latencies revealed that target fixations peaked later with No-Delay (dotted lines, square) relative to Long- (solid lines, circle) and Short-Delays (dashed lines, triangle). The timing of long and short delays did not significantly differ. Curves represent observed data, dots represent peak latency, and horizontal lines represent standard error.

Figure 9.

Competitor (black) and control (gray) fixations in the No-Load (Experiment 1), Long-Delay (Experiment 2), and Short-Delay conditions (Experiment 2). Effects of competition on peak latency increased with shorter delays between the target preview and search display onset, with the largest competitor lead in the Short-Delay condition (delay: 250ms), followed by the Long-Delay condition (delay: 750ms), and then the No-Load condition (delay: 1000ms). Curves represent observed data, dots represent peak latency, and horizontal lines represent standard error.