Contributions from Long-Term Memory Explain Superior Visual Working Memory Performance with Meaningful Objects

Abstract

Visual working memory (WM) capacity has been claimed to be larger for meaningful objects than for simple features, possibly because richer semantic representations enhance the distinctiveness of stored items. However, prior demonstrations typically compared trial-unique meaningful objects with a small set of repeated simple features. This design confounds meaningfulness with proactive interference (PI), such that PI is minimal for trial-unique objects but substantial for repeated features. As a result, superior performance for meaningful objects may reflect contributions from episodic long-term memory (LTM) rather than expanded WM capacity. To test this, Experiment 1 measured WM for repeated colors, repeated meaningful objects, and trial-unique meaningful objects. The advantage for objects over colors was replicated in the trial-unique condition, but eliminated for repeated objects that equated PI across stimulus types. Hierarchical Bayesian dual-process modeling revealed that the trial-unique advantage reflected stronger familiarity signals, whereas recollection remained stable across stimulus types. Experiment 2 assessed WM storage directly using contralateral delay activity (CDA), an electrophysiological marker of the number of items stored. Although trial-unique objects again yielded behavioral advantages, CDA activity across increasing set sizes revealed a common slope and plateau for trial-unique meaningful objects and repeated colors. The CDA difference between stimulus types was additive and did not vary with set size, providing no evidence for an increased number of stored items. These findings demonstrate that previously reported advantages for meaningful objects primarily reflect reduced PI and enhanced LTM familiarity. When PI is equated, WM storage limits for simple and meaningful stimuli are equivalent.

Figure 1.. Mean contralateral delay activity (CDA) amplitudes for color, orientation, and conjunction stimuli at set sizes 2 and 4, adapted from Figure 3D of Woodman and Vogel (2008).

Although orientation and conjunction stimuli elicited larger CDA amplitudes than color stimuli, these differences were additive across set sizes, suggesting that the enhanced CDA was not due to an increased number of items stored in working memory.

Figure 2.. Experimental paradigm and behavioral results from Experiment 1.

(A) Task sequence for the visual WM recognition task. Each trial began with a fixation display (500 ms), followed by a memory array containing six items (1,000 ms), a blank retention interval (1,500 ms), and a recognition probe requiring a confidence rating on a continuous scale ranging from “sure new” to “sure old”. (B) Examples of stimulus sets used across three conditions: colors (CLR_REP; repeated across trials), meaningful objects repeated throughout the experiment (OBJ_REP), and trial-unique meaningful objects presented only once (OBJ_UNI). (C) Behavioral performance summarized as hit rate (left), false-alarm rate (middle), and estimated visual working memory capacity measured by Cowan’s K (right) for each stimulus condition. Error bars represent standard error of the means.

Figure 3.. Observed and model-fitted ROC curves for individual participants.

Each panel shows hit rate (HR) plotted over false-alarm rate (FAR) across confidence levels from each participant. Solid curves represent model predictions generated by the mean posterior parameters estimated from the hierarchical Bayesian dual-process signal detection model. Color codes represented repeated-colors (CLR_REP in red), repeated-objects (OB_JREP in green), and unique-objects (OB_JUNI in blue) conditions, respectively. Circles denote observed data. The close correspondence between predicted and observed values indicates good individual-level model fit.

Figure 4.. Observed and model-predicted ROC curves and posterior parameter estimates from the hierarchical Bayesian DPSD model.

(A) ROC curves for each stimulus condition: repeated colors (CLR_REP; red), repeated meaningful objects (OBJ_REP; green), and trial-unique meaningful objects (OBJ_UNI; blue). Circles represent observed mean hit rates (HR) and false-alarm rates (FAR) across confidence levels, and the horizontal and vertical error bars indicate standard errors of the mean FAR and HR, respectively. Solid lines depict model-predicted ROC curves based on posterior mean values of the model parameters. (B) Posterior means and 95% highest density intervals (HDIs_95%) for the population-level recollection (left) and familiarity (right) parameters, across stimulus types. The boundaries of HDI_95% not crossing over between conditions indicate a statistically credible difference.

Figure 5.. Hypothetical CDA amplitude patterns predicted by competing hypotheses.

(A) If remembering meaningful objects truly expands visual WM capacity, the CDA amplitude should exhibit a significant set size × stimulus type interaction effect, with CDA amplitude increasing across a larger range of set sizes (set size 5). (B) By contrast, a stimulus-related confound hypothesis predicts only additive amplitude differences between stimulus types across all set sizes, without interaction.

Figure 6.. Procedure and resulting capacity estimates for Experiment 2.

(A) Lateralized visual WM task sequence shown for the color stimulus condition. Each trial began with central fixation (500 ms), followed by an arrow cue indicating the task-relevant hemifield (left or right; 500 ms). The memory array was then presented laterally (1,000 ms), followed by a blank retention interval (1,500 ms). Participants subsequently indicated whether a test probe matched the remembered item at the corresponding location. The fixation point changed its color to white to indicate the onset of test array and allow participants to blink. The trial-unique meaningful object condition used the same procedure, differing only in stimulus content. (B) Capacity estimates of mean Cowan’s K, as a function of set size (1, 3, or 5 items) and stimulus type (unique object vs. repeated color). Error bars represent standard error of the mean.

Figure 7.. Grand-averaged contralateral and ipsilateral waveforms by stimulus type and set size.

Waveforms are time-locked to the memory array onset (0 ms) and offset (1000 ms) and averaged across posterior electrode sites (P3/P4, P7/P8, PO3/PO4, PO7/PO8) in Experiment 2. Each panel compares contralateral (red) and ipsilateral (black) activity for trial-repeated colors (top row) versus trial-unique objects (bottom row) at set sizes 1, 3, and 5 (left to right). Vertical dotted lines mark the onset and offset of the memory array.

Figure 8.. Grand-averaged CDA waveforms and mean amplitudes for trial-repeated colors and trial-unique objects.

(A) CDA activity (contralateral minus ipsilateral activity at posterior electrode sites; P3/P4, P7/P8, PO3/PO4, PO7/PO8) time-locked to the memory onset and averaged across participants, plotted separately for repeated colors (left) and unique objects (right) at set sizes 1 (green), 3 (sky blue), and 5 (blue). Vertical dotted lines mark stimulus onset (0 ms) and offset (1000 ms), and gray shading indicates Encoding (400–1000 ms from stimulus onset), Delay (1400–2000 ms), and Combined (400–2000 ms) measurement windows. Shaded error bars represent standard error of the means (SEM). (B) Mean CDA amplitudes during the early (left), late (middle), and combined (right) time windows as a function of set size for unique objects (triangles on blue line) versus repeated colors (circles on red line). Error bars represent SEM.