Working memory updating of emotional stimuli predicts emotional intelligence in females

The study examines whether the efficiency of working memory updating (WMU) differs in its relation to general intelligence (gF/g) versus emotional intelligence (EI), and whether these relations depend on the emotional content of the material being processed ('hot' emotional vs. 'cool' neutral). Prior work shows strong links between working memory capacity (WMC) and fluid intelligence, both considered domain-general mechanisms operating across tasks. EI, conceptualized as an ability (per Mayer-Salovey), concerns processing affectively meaningful information and has shown positive relations with general intelligence when measured by performance-based tests. However, the specific cognitive processes underlying EI are less clear, and prior studies combining multiple processes yielded inconclusive results. This study focuses specifically on WMU as a key executive function, hypothesizing that while IQ relates to WMU regardless of content, EI depends particularly on WMU for emotional stimuli. The purpose is to test domain specificity (EI) versus domain generality (IQ) using parallel neutral and emotional WMU tasks and structural equation modeling (SEM).

Extensive literature links WMC/WM processes to gF and broader cognitive skills (reading, verbal comprehension, mathematics, problem solving). gF is defined as a domain-general capacity, suggesting its relations to WM should be content-independent. EI, in contrast, is 'hot' intelligence involving emotion-laden, personally relevant information; ability-based EI measures typically correlate positively with g, unlike many self-report EI questionnaires. Preliminary evidence (e.g., Gutiérrez-Cobo et al.) indicates higher EI associates with better performance on WM tasks with emotional (hot) content but not necessarily neutral content. WMU has been identified as a central executive function strongly related to gF and predictive of various academic and cognitive outcomes. Evidence also suggests that removing irrelevant negative WM content relates to emotion regulation in social contexts. Together, these findings motivate testing whether EI specifically relates to WMU for emotional material, whereas IQ relates to WMU irrespective of content.

Participants: 123 women (Mage=39.71, SD=8.67) completed all measures except the memory updating task (MUT) in session one; a subset of 69 women completed the MUT in a second session. Informed consent was obtained. Ethics approval was granted by SWPS University; procedures followed the Helsinki Declaration. Measures: - Fluid intelligence (gF): Raven’s Advanced Progressive Matrices (RAPM), 40-minute limit. - Crystallized intelligence (gC): Horn and Cattell-based verbal classification test (120 words categorized into art, biology, science, literature, geography/history) completed in 5 minutes; score is number correct. - Emotional intelligence (EI): TIE ability test based on Mayer-Salovey model with four branches (perception, facilitation, understanding, managing). 24 tasks (6 per branch). Participants rate three responses per scenario on a 5-point scale. Scoring is similarity to expert judges (52 psychologists). Reliability r=0.88. In prior data, TIE correlated with RAPM r=0.35 and Horn r=0.26 (p<0.001). Memory Updating Task (MUT): Original computerized task assessing WMU efficiency. Two parallel blocks: neutral content (objects) and emotional content (human faces with emotions). Each trial presents four sequential pairs of pictures (fixed set size). For each pair: pictures displayed 2000 ms, blank 1000 ms, then a red frame (500 ms) appears at the location of one picture, indicating the target to remember. Initially, both items must be encoded; WM updating occurs when the frame appears, designating the target and rendering the other item a distractor. Recognition phase: after the sequence, participants are shown a pair number (1–4) and a picture and decide whether the picture was the indicated item for that pair (yes/no). Content manipulation: neutral vs. emotional. Similarity manipulation: low vs. high (neutral: different vs. same category objects; emotional: different people vs. same person). Distractor types (3): (1) unindicated picture from the correct pair; (2) indicated picture from another pair; (3) picture not shown in the series. Design: 2 (content) × 2 (similarity) × 4 (response types: positive vs. three incorrect types). Total 96 trials evenly distributed; two blocks (neutral, emotional), block order counterbalanced; within-block conditions randomized. Accuracy and reaction times recorded. Emotional faces from NimStim; neutral objects from free stock; sets balanced on physical characteristics and differed in valence. Procedure: Session 1 (paper-pencil): RAPM (40 min), Horn/Cattell (5 min), TIE (30 min). Session 2 (~45 min): MUT neutral and emotional versions, order randomized. Statistical analysis: Structural equation modeling (SEM) in AMOS 21 with maximum likelihood and full information maximum likelihood (FIML) for missing data. Model fit assessed by χ² (nonsignificant desirable), CFI (>0.95), RMSEA (<0.06), SRMR (<0.08). WMU indicators were percent correct in neutral and emotional MUT blocks. Measurement model specified two latent factors: EI (indicated by TIE branches) and IQ (indicated by RAPM and Horn), with their covariance. Structural paths from MUT emotional and MUT neutral to EI and IQ were then estimated. Residual covariance between MUT indicators was allowed due to shared task characteristics.

Descriptive and correlations (Table 1): - TIE general M=25.64, SD=6.24, N=123. - RAPM M=12.43, SD=5.49, N=121; Horn M=36.01, SD=18.98, N=120. - MUT neutral accuracy M=0.61, SD=0.13, N=69; MUT emotional accuracy M=0.57, SD=0.09, N=68. - TIE general correlated with RAPM r=0.28 (p<0.05) and Horn r=0.53 (p<0.05). TIE branches showed significant intercorrelations and with IQ measures. MUT emotional showed stronger relations with TIE scores than MUT neutral. Both IQ measures positively related to both MUT versions. SEM results: - Measurement model for EI and IQ: good fit, χ²=12.06, p=0.149; CFI=0.985; RMSEA=0.064. Loadings: EI indicated by TIE Perception (β=0.73), Understanding (β=0.83), Facilitation (β=0.81), Managing (β=0.68), all p<0.001. IQ indicated by RAPM (β=0.50) and Horn (β=0.92), p<0.001. EI–IQ covariance r=0.57, p<0.001. - Structural relations: • MUT neutral → IQ: β=0.34, p=0.008 (significant). • MUT neutral → EI: β=0.003, p=0.980 (ns). • MUT emotional → EI: β=0.53, p<0.001 (significant). • MUT emotional → IQ: β=0.26, p=0.044 (significant but weaker than to EI). - Residual covariance between MUT indicators r=0.49, p<0.001. - Constraining the nonsignificant MUT neutral → EI path to zero did not worsen fit (Δχ²₁=0.01, p=0.980). Final model fit: χ²₁₆=19.93, p=0.223; CFI=0.990; RMSEA=0.038. Interpretation: In females, WMU for emotional stimuli robustly predicts EI and also relates to IQ; WMU for neutral stimuli predicts IQ but not EI. This pattern supports domain specificity of EI with respect to emotional WM updating and domain generality of IQ with respect to WM updating across content types.

Findings indicate that while EI and IQ are positively related when EI is measured by an ability test, they diverge in how they relate to WM updating. IQ relates to WMU irrespective of stimulus content, consistent with its domain-general status and established links between WM and gF/gC. EI, however, shows a content-specific relation: only WMU for emotional stimuli predicts EI. This suggests that efficiently updating emotion-laden information—retaining relevant affective cues and suppressing irrelevant ones—underpins emotionally intelligent behavior. The stronger association of EI with crystallized intelligence than with fluid intelligence also aligns with prior research on ability EI tests. Overall, EI appears more domain-specific in its cognitive underpinnings, whereas IQ reflects domain-general WM processes. These results extend prior preliminary evidence on hot vs. cool WM-EI links and clarify WMU’s role as a key mechanism for EI in females.

The study contributes evidence that, in females, the efficacy of working memory updating for emotional material is a strong predictor of ability-based emotional intelligence, whereas working memory updating for neutral material does not predict EI. In contrast, both emotional and neutral WM updating predict IQ, supporting the domain-general relationship between WM and IQ. The work clarifies the domain-specific cognitive mechanism (emotional WM updating) underlying EI and reinforces WMU’s central role in higher cognition. Future research should replicate these findings in larger, gender-balanced samples, employ diverse measures of IQ, EI, and WMU, and examine whether similar patterns hold for males and across different emotional contexts and task paradigms.

Primary limitations include relatively small sample sizes (n=123 for IQ/EI measures; n=69 for WMU), female-only participants limiting generalizability to women, and potential measure-specific effects since single measures of IQ, EI, and WMU were used. Although SEM indicators showed strong loadings and acceptable distributional properties, the sample size is near lower bounds for SEM. Replication with larger, mixed-gender samples and alternative instruments is recommended to assess robustness and generality.