Value network engagement and effects of memory-related processing during encoding and retrieval of value

Abstract

Decision makers rely on episodic memory to calculate choice values in everyday life, yet it is unclear how neural mechanisms of valuation differ when value-related information is encoded versus retrieved from episodic memory. The current fMRI study compared neural correlates of value while information was encoded versus retrieved from memory. Scanned tasks were followed by a behavioral episodic memory test for item-attribute associations. Our analyses sought to (i) identify neural correlates of value that were distinct and common across encoding and retrieval, and (ii) determine whether neural mechanisms of valuation and episodic memory interact. The study yielded three primary findings. First, value-related activation in the fronto-striatal reward circuit and posterior parietal cortex was comparable across valuation phases. Second, value-related activation in select fronto-parietal and salience regions was significantly greater at value retrieval than encoding. Third, there was no interaction between neural correlates of valuation and episodic memory. Taken with prior research, the present study indicates that fronto-parietal and salience regions play a key role in retrieval-dependent valuation and context-specific effects likely determine whether neural correlates of value interact with episodic memory.

Keywords: Decision making; Encoding; Episodic memory; Retrieval; Value network; fMRI.

Figure 1. Brain regions implicated in valuation

Note. Previous research has implicated a variety of brain areas spanning the fronto-striatal (in light blue), fronto-parietal (in dark blue), and salience networks (in red) in different aspects of value processing. Abbreviations: OFC – orbitofrontal cortex; vmPFC – ventromedial prefrontal cortex; VS – ventral striatum; dACC – dorsal anterior cingulate cortex; Thal – thalamus; PCC – posterior cingulate cortex; Precun – precuneus.

Figure 2. Visualization of study goals

Note. (A)) The primary study goal was to identify distinct and common neural mechanisms of valuation across value encoding and retrieval. To address this goal, brain activation associated with value for consumer products was determined while value-related information was present (value-encoding phase), and later, when value-related information was absent (value-retrieval phase). Resulting phase-subtraction and conjunction-based ROI masks were used to determine whether value-related activity was significantly distinctive during value encoding and retrieval. (B) The secondary study goal was to test for evidence of interaction versus independence between neural mechanisms of valuation and episodic memory. In the case of interaction, we then aimed to determine the nature of the relationship between these systems by probing for competition versus cooperation. To address this goal, demands on memory was manipulated at encoding and retrieval, with corresponding neuroimaging analyses examining effects of encoding experiences on value-related brain activation. Additional analyses were used to determine whether brain activation related to value differed with successful encoding of associated episodic details (product attributes). An absence of memory-dependent effects on the neural correlates of value was taken as evidence for independence. Enhanced value-related activation with more encoding experiences or successful episodic encoding was taken as evidence of cooperation (reciprocal green arrows). Diminished value-related activation with more encoding experiences or successful episodic encoding was taken as evidence of competition (inhibiting red lines).

Figure 3. Consumer Judgment Task design

Note. The Consumer Judgment Task was divided into three phases: value encoding, value retrieval, and attribute memory. Value encoding and retrieval phases were completed during fMRI; the attribute memory task followed outside the scanner. (A) During value encoding, reviews of consumer products were rated using a discrete 1 (negative) to 4 (positive) value scale. Products were either presented with one consumer review or two different consumer reviews. During value retrieval, the average rating for each product was estimated based on the previously encountered consumer reviews, using a continuous 1 (negative) to 5 (positive) rating scale. (B) Attribute memory was tested using a 1 (definitely yes) to 4 (definitely not) scale as to whether attribute-product pairs were presented together during encoding.

Figure 4. Correspondence between value ratings across value encoding and retrieval

Note. Normalized value ratings during value retrieval were found to scale alongside value encoding ratings (averaged for products with two consumer reviews), regardless of the number of encoding experiences. For visual comparison purposes, discrete value encoding ratings (X-axis) and continuous value retrieval ratings (Y-axis) were normalized to a 0-1 scale, representing negative (0) to positive (1) value ratings. Shading represents 95% confidence intervals for each regression line.

Figure 5. Phase-distinct and common neural activity related to value ratings

Note. (A) Greater overall recruitment of fronto-striatal valuation and fronto-parietal control network areas was associated with encoding and integrating value information, with areas of the bilateral striatal caudate nucleus and left superior parietal lobule surviving FDR-correction (voxel-wise threshold of q = .05). (B) While retrieving value information from recent memory, robust neural recruitment was primarily observed in salience and fronto-parietal control network areas, including superior frontal, insular, dorsal anterior cingulate, precuneal, and superior parietal clusters. (C) Recruitment of fronto-parietal and sensory processing regions was associated with value processing across valuation phases, with clusters in the left superior parietal lobule, precuneus, and right occipital fusiform areas surviving FDR-correction. Image slices are presented in radiological orientation with MNI slice coordinates.

Figure 6. Comparison of ROI cluster activity levels across valuation phases

Note. (A) Similar value-related activation levels were observed across valuation phases within the caudal (light blue) and superior parietal (dark blue) clusters identified from the value encoding subtraction contrast, reflecting dependence on striatal and posterior parietal areas for value processing whether the value details were present or retrieved from recent memory. (B) During the value-retrieval phase, however, significantly greater activity was observed for the dorsal anterior cingulate gyrus (red; t₍₁₉₎ = −2.26, p = .04, d = −.51), left lateralized insular cortex (red; t₍₁₉₎ = −2.75, p = .01, d = −.62), and precuneus (dark blue; t₍₁₉₎ = −2.75, p = .01, d = −.61) in comparison to activity at value encoding. (C) Posterior parietal ROIs (dark blue) identified from a conjunction analysis represent regions recruited for value processing across phases and therefore were not expected to have significant activation differences. Error bars represent standard error of the mean; asterisks (*) indicate significant differences in activity between phases at p < .05.