As can be seen in , the CHI and controls groups were well matched in age, education, estimated premorbid IQ, and sex ratio. The control group contained a larger proportion of minority participants than the CHI group. However, the controls outperformed the CHI group on the CVLT (see below). Hours of posttraumatic amnesia (PTA), years of age, and years post injury were not significantly correlated with CVLT Trials 1–5 Recall (r
=−.19, −.21, −.05, respectively; p
s> .13), Short-Delay Free Recall (r
=−.18, −.12, .08, respectively; p
s> .23), nor Long-Delay Free Recall (r
=−.23, −.14, .05, respectively; p
s> .14) in our CHI group, consistent with the work of Chu et al. (2007)
who showed that PTA and age at injury no longer predicted total learning performances on the Rey Auditory Verbal Learning Test beyond one year post injury via mixed-effects modeling in a longitudinal study including a sample of 794 traumatic brain injury survivors.
The CHI group performed significantly worse than the controls across all of the major indices of the CVLT (see ), with effect sizes ranging from medium to large. Additionally, the CHI group exhibited greater deficits on all of the ISDA indices than the controls, with the group difference on the encoding deficit index representing a large effect and the group differences on the consolidation and retrieval deficit indices indicating medium effects. It should be noted that acquisition-corrected values were used for the consolidation and retrieval deficit indices to control for CHI-related encoding difficulties.
Impact of Memory Process Deficits
Acquisition-corrected values were not used in the regression analyses, since no group comparisons were made. The raw ISDA indices were not strongly associated (rs<.44). The predictors were ordered in the regression by their natural temporal relationship (i.e., 1. encoding, 2. consolidation, 3. retrieval). The encoding index significantly predicted long-delayed free recall (Model 1), FΔ(1, 54)=93.42, accounting for 63% of the variance (see ). The consolidation index added significant predictive power to the model (Model 2), FΔ(1, 53)=32.39, accounting for an additional 14% of the variance. The retrieval index added marginal predictive power to the model (see Model 3), FΔ(1, 52)=5.19, accounting for an additional 2% of the variance. Overall, the ISDA encoding and consolidation indices (Model 2) were the most notable predictors of delayed recall in the CHI group.
Summary of Hierarchical Regression to Determine the Impact of CHI Associated Encoding, Consolidation, and Retrieval Deficits on Long-Delay Free Recall
Role of Clustering in Encoding Deficits
To determine any between- or within-group differences in serial clustering over learning trials, a 2 (group) × 5 (learning trial) ANOVA, with repeated measures on the last factor, was conducted. No effect of group, trial, or group × trial interaction was detected, Fs≤ 2.34. These findings indicate that CHI participants (trial 1, M=.33, SD=.77; trial 2, M=.34, SD=.97; trial 3, M=.41, SD=.98; trial 4, M=.48, SD=1.22; trial 5, M=.80, SD=1.40) and controls (trial 1, M=.14, SD=.74; trial 2, M=.26, SD=.96; trial 3, M=.27, SD=1.22; trial 4, M=.31, SD=1.31; trial 5, M=.33, SD=1.35) demonstrated similar patterns of serial clustering over the learning trials.
The same analysis was completed for semantic clustering and this revealed a significant effect of group, F(1,116)=7.29, MSe=24.05, ηp2=.06; trial, F(4,116)=39.09, MSe=2.43, ηp2=.25; and a significant group × trial interaction, F(4,116)=2.99, MSe=2.43, ηp2=.03. Follow-up paired samples t-tests (Bonferroni correction: α < .0125) revealed that the controls showed significantly improved semantic clustering between trials 1 and 2, (trial 1, M=.90, SD=1.71; trial 2, M=1.63, SD=2.60), t(1,61)=−3.02, d=.33; trials 2 and 3, (trial 3, M=2.20, SD=2.96), t(1,61)=−2.64, d=.20; and 3 and 4, (trial 4, M=3.41, SD=3.31), t(1,61)=−4.31, d=.39, but their clustering remained stable between trials 4 and 5 (trial 5, M=3.53, SD=3.44), t(1,61)=−0.46, d=.04. Conversely, the CHI group showed a significant improvement between trials 3 and 4, (trial 3, M=1.02, SD=1.70; trial 4, M=1.93, SD=2.85), t(1,55)=−2.98, d=.39, but not between any other contiguous trials, ts≤−1.69, ds≤.18 (trial 1, M=.57, SD=1.31; trial 2, M=.70, SD=1.79; trial 5, M=1.98, SD=3.09). These data indicate that the CHI participants, in comparison to controls, suffered from impairments in the ability to initiate and utilize semantic clustering strategies during list learning (see ).
Mean semantic clustering (± SE) by learning trial for the CHI and control groups.
Finally, multiple linear regression was used to determine if 1) average semantic clustering across trials, 2) changes in semantic clustering over trials, or 3) the interaction between the two variables predicted encoding deficits in the CHI group. The model was significant, R2 =.48, F(3,52)=15.73, although only average semantic clustering emerged as a significant predictor, B=−1.19, t=−3.27; change in clustering, B=.02, t=−.09, and average clustering × change in clustering, B=.01, t=−.09, were not significant predictors. It should be noted that average semantic clustering and change in semantic clustering were not intercorrelated (r=0.03). Overall, these results suggest that reduced semantic clustering plays a large role in encoding deficits for verbal material in individuals with severe CHI.
Impact of Encoding on Consolidation
To determine if greater encoding benefited consolidation of list items, ISDA consolidation deficit indices were calculated for both high acquisition (items recalled 4 or 5 times during learning) and low acquisition (items recalled 1 or 2 times during learning); these values reflect the median split of measurable list acquisition, as the maximum recall value for a given item on the CVLT is 5. A 2 (group) × 2 (acquisition level) ANOVA, with repeated measures on the last factor, revealed significant effects of group, F(1,116)=4.43, MSe=0.06, ηp2=.04, and acquisition level, F(1,116)=80.03, MSe=.04, ηp2=.41, but no group × acquisition level interaction, F<1. These results indicated group differences in consolidation across acquisition levels and greater consolidation deficits for items acquired at a low level (CHI: M=0.34, SD=0.30, control: M=0.27, SD=0.29) as opposed to a high level (CHI: M=0.10, SD=0.15, control: M=0.04, SD=0.07) (see ). Overall, these findings suggest that level of acquisition plays a role in consolidation difficulties, but like the hierarchical regression results presented above, that CHI survivors exhibit consolidation difficulties above and beyond their encoding deficits.
Mean ISDA consolidation deficit index values (± SE) for high (items recalled 4 or 5 times during learning) and low (items recalled 1 or 2 times during learning) acquisition levels for the CHI and control groups.