,. Next we validated that increasing similarity between the colors of objects increased the competition (interference) between them. Indeed objects presented in highly similar colors were harder to discriminate as evidenced by a steeper learning curve for these pairs. Turning toward the color memory test, color error for the objects steadily decreased over learning in all similarity conditions suggesting participants were able to learn each object's associated color.
Next we turned to the critical post test to assess if and how color memory for the objects was distorted over learning. After completing all 8 rounds of study, color error in the post-test marginally varied across similarity conditions (p=0.073). Color error was lower for moderate similarity objects compared to low similarity (p=0.037) and high similarity color error did not differ from either moderate (p=0 .62) or low similarity (p-0.13) color error. However, we were most interested in whether there was a bias in color errors relative to the competing color. We quantified the bias in color estimates two ways. First, we signed the distance of each error depending on whether the color estimate was positioned towards the competing color (positive error) or away from the competing color (negative error). Thus, any deviation in this measure from zero would indicate a bias in color memory relative to the competing color with positive values reflecting an attraction towards and negative values reflecting a repulsion away from the competing color. Second, we calculated the percentage of trials in which the estimate fell away from the competing color. Similarly, any deviations in the measure from 50% would reflect a bias in color memory. While signing the error allowed for an estimate of the effect size (in degrees) of the distortion of the competing colors in color space, the percent error estimate was more robust to outliers in color memory as this measure was less influenced by extreme values in color error. Critically, both measures of color bias scores highly varied by similarity condition (percent away: p=0.0002, signed error: p=0.0004). While there was a strong color bias away from the competing color in the high similarity condition (percent away: signed error: p=0.002), there was no bias in either the moderate (percent away: p=0.33 , signed error: p=0.22) or low similarity (percent away: p=0.92, signed error: p= 0.20) conditions. Thus participants systematically exaggerated the color of the object away form the competing color when competition between the colors was highest. This effect persisted after a 24 hour delay as there remained a selective color bias in the high similarity condition on the Day2 post-test.
Next we tested the extent to which this repulsion bias in color estimates was adaptive to memory performance. Specifically we asked if the degree of exaggeration between the competing colors in the high similarity condition was related to improved discrimination performance between their associated objects. Indeed, we found a positive correlation between the extent of color bias in the post-test and performance at discriminating between the objects during the association test. Thus subjects who showed greater repulsion of the similar colors in memory demonstrated less interference between those corresponding objects.
Feature repulsion depends on task demands
Next we asked wether the repulsion of similar color values depends on the task demand to separate the objects or if it is a byproduct of any associative learning. In the previous experiments, the associative memory task required participants to distinguish between the two similar colored objects. Here, Experiment 3 was run identically to Experiment 1, expect we altered the task during learning so that optimal performance required integrating across the overlapping face-object pairs. Specifically, during the test, one of the learned faces appeared and participants had to select the face (from a set of 4 faces) that was paired with the same object. With this subtle task change, discriminating the color of the objects was no longer relevant to performance on the associative task. Rather, optimal performance required linking between the two faces in memory via their shared association with the common object (irrespective of its color). We then tested how this change in task demand - from separating the colors to generalizing across them - would effect the repulsion bias. Would the bias remain, diminish, or possibly flip resulting in an attraction bias reflecting a merging of the color values?
Participants successfully learned to infer the face pairings as they chose the target face at rates well above chance in all conditions (ps <0.00001). Color error also decreased over learning in all conditions.
After 8 rounds of learning color error varied by similarity condition (p = 0.003). Follow up t-tests revealed that there were larger color errors for objects in the low similarity condition compared to the both the high (p = 0.007) and the moderate (p=0.025) similarity conditions. However, color error did not differ between the high and moderate similarity conditions (p=0.11). Turning to our main question of interest we next investigated if the change in task demand altered the bias in color estimates. As in Experiment 1 both measures of color bias varied by similarity condition (percent away: p=0.015; signed error: p = 0.002), however, in this experiment there was no color bias in the high similarity condition (percent away: p = 0.56, signed error: p = 0.29). Surprisingly, both the moderate (percent away: p=0.06; signed error: p = 0.007) and high (percent away: p=0.00004; signed error: p = 0.00009) similarity condition showed positive bias towards the competing color. Direct t-test comparisons of the bias measures across Experiment 1 and Experiment 3 revealed a significant difference in color bias for the high similarity (percent away: p = 0.001; signed error: p=0.009)and low similarity (percent away: p = 0.001; signed error: p = 0.002) conditions but not for the moderate similarity condition (percent away: p = 0.11; signed error: p = 0.38). Thus, changing the task demand to encourage integration eliminated the repulsion bias in the high similarity condition and induced an attraction bias in the low similarity condition.
The limits of repulsion
Last, we explored how
In this study we created an ultra similarity condition in which the colors of the object pairs were only separated by 6 degrees. In addition to the ultra similarity condition we kept the high and moderate similarity conditions as comparison points. It also gave us another opportunity to replicate the repulsion bias in the high similarity condition found in Experiments 1 and 2.
local maximum at 24,
24 degrees seem to be relevant sweet spot
Discussion
Summary
however, to date there have been no studies examining how the features of those corresponding memories change over learning. Using a novel behavioral paradigm that assessed memory on a continuous scale, we show that learning induced a repulsion bias in color memory between competing memories. Whereas color memory was biased towards the competing color early in learning, it was biased away from the competing color by the end of learning. This repulsion bias was competition dependent as it was only observed for colors that were highly similar (24 degrees apart). Furthermore, the repulsion of competing colors was adaptive to memory performance as greater repulsion between the colors was associated with reduced interference between them.
The repulsion bias observed for the features of memory, strongly parallels recent fMRI findings that initial overlap in hippocampal patterns triggers of repulsion of their patterns with learning. In both these studies and the fMRI Future work will need to test if the two results are related.
design features
1)
only found in the high similarity condition
Requires task demand to repel