Overall, accuracy was very high, above 85%, in all conditions (Fig. 2). A 2 × 2 repeated measures ANOVA with novelty (novel vs repeating) and retrieval probe type (match vs nonmatch) revealed only a main effect of probe type (F(1, 18) = 14.06, p 0.002, ηp2 = 0.44), with matching probes having a mean accuracy of 87.9% ± 2.9% Standard Error (SE) and non-matching probes having a mean accuracy of 95.4% ± 1.9% SE. All other effects were non-significant (all p > 0.05).

Averaged accuracy averaged over three runs, separately for matching and non-matching trials. Overall, accuracy was very high (above 85%) in all conditions: novel sequence, match; repeated sequence, match; novel sequence, non-match; repeated sequence, non-match. A significant main effect of probe type emerged, with matching probes having a mean accuracy of 87.9% ± 2.9% SE and non-matching probes having a mean accuracy of 95.4% ± 1.9% SE. Bars represent the standard error. *p = 0.002

Individual median reaction time in milliseconds on correct trials was averaged across the three runs and calculated separately for novel vs repeating trial types, and matching vs non-matching retrieval probes. Results for these trial types are presented in Fig. 3.

Averaged median reaction time for correct trials in milliseconds averaged over three runs, separately for matching and non-matching trials. Shown are the results of the conditions: novel sequence, match; repeated sequence, match; novel sequence, non-match; repeated sequence, non-match. A significant main effect of probe type was evident, with reaction times for matching probes (852.3 ms ± 52.7 ms SE) significantly lower than those for non-matching probes (928.7 ms ± 59.8 ms SE). Bars represent the standard error. *p = 0.004

A 2 × 2 repeated measures ANOVA with novelty (novel vs repeating) and retrieval probe type (match vs nonmatch) revealed only a significant main effect of probe type (F(1, 18) = 10.70, p = 0.004, ηp2 = 0.37). Reaction times for matching probes (852.3 ms ± 52.7 ms SE) were significantly lower than those for non-matching probes (928.7 ms ± 59.8 ms SE). All other effects were non-significant (all p > 0.05).

Figure 4A–C shows the MNI-normalized MPRAGE volume averaged across all subjects and sliced sagitally at the X = – 1, – 3, and – 5-mm coordinate (left side). The localization of the inferior olive ROI from the Brainstem Nuclei Atlas has been enhanced in these sections by artificially brightening the region. The location depicted corresponds well with the structural appearance of the inferior olive seen on histologically stained sections (Fig. 4D).

Inferior Olive ROI of the Brainstem Nuclei Atlas. MNI-normalized MPRAGE volume averaged across all subjects and sliced sagitally at the X = – 1, – 3, and – 5-mm coordinate (A, B, C, respectively). The localization of the inferior olive ROI from the Brainstem Nuclei Atlas has been enhanced in these sections by artificially brightening the region. D Structural appearance of the inferior olive (IO) seen on histologically stained sections. The location depicted in A–C corresponds well with the structural appearance in D Adapted with permission from https://brainmuseum.org/, Specimens used for this publication are from the Defense Health Agency Neuroanatomical Collections Division of the National Museum of Health and Medicine, the University of Wisconsin and Michigan State Comparative Mammalian Brain Collections supported by the US National Science Foundation

Repeated measured ANOVA revealed a significant main effect of phase on inferior olive activation (F(3, 54) = 10.51, p < 0.001, ηp2 = 0.37), no main effect of novelty alone (F(1, 18) = 0.07, p = 0.8, ηp2 < 0.01), and a significant interaction effect of phase x novelty on inferior olive activation (F(3, 54) = 3.53, p = 0.021, ηp2 = 0.16). Average left inferior olive activation during encoding, maintenance, and retrieval for novel and repeating sequences is presented in Fig. 5.

Average inferior olive activation during encoding, maintenance, and retrieval (match and non-match). The average BOLD activation of the left inferior olive is shown during each phase for repeated and novel letter sequences. Activation is evident during encoding for novel sequences (M = 1.30, SE = 0.43) and repeating sequences (M = 0.69, SE = 0.27). Relative to encoding, reduced activation is visible during maintenance for both novel (M = – 0.71, SE = 0.33) and repeating (M = – 0.12, SE = 0.35) sequences. Increased activation was evident during retrieval for novel matching (M = 2.37, SE = 0.59), novel non-matching (M = 1.20, SE = 0.56), repeating non-matching (M = 2.62, SE = 0.62) with lowest activation seen during retrieval for the repeating matching condition (M = 0.73, SE = 0.56). Bars represent the standard error

Activation is evident during encoding for novel sequences (M = 1.30, SE = 0.43; one sample t-test: t(18) = 3.01, p = 0.007, d = 0.7) and repeating sequences (M = 0.69, SE = 0.27; t(18) = 2.60, p = 0.018, d = 0.6). The predicted reduction in encoding activation for repeating sequences relative to novel sequences approached significance (F(1, 18) = 3.77, p = 0.068, ηp2 = 0.17). Relative to encoding, reduced activation is visible during maintenance for both novel (M = – 0.71, SE = 0.33) and repeating (M = – 0.12, SE = 0.35) sequences. The encoding-to-maintenance reduction was significant for novel sequences (F(1, 18) = 11.20, p = 0.004, ηp2 = 0. 38), and approached significance for repeating sequences (F(1,18) = 3.09, p = 0.096, ηp2 = 0.15). The reduction in activation from encoding to maintenance for novel sequences was significantly greater than that observed for repeating sequences (F(1, 18) = 4.53, p = 0.047, ηp2 = 0.20).

Considering the maintenance phase alone, the activation for repeating sequences was not significantly different from 0 (t(18) = – 0.34, p = 0.738, d = – 0.08), whereas the maintenance phase activation for novel sequences went significantly below 0 (t(18) = – 2.19, p = 0.042, d = – 0.5). However, direct comparison of novel vs. repeated maintenance phase activation did not reach significance (F(1, 18) = 2.87, p = 0.107, ηp2 = 0.14); see Fig. 6.

Visual presentation of the three contrast analyses performed for phase and novelty. Shown are examples of average inferior olive activation in the respective phase and the anatomical location of the inferior olive according to the Brainstem Nuclei Atlas from the Brainstem Imaging Laboratory. Red indicates average activation in the individuals. En Encoding novel sequences, Mn Maintenance novel sequences, Mr Maintenance repeated sequences, Rrn Retrieval repeated sequences, non-matching probes. The threshold for the activation maps is set to p = 0.005 for the retrieval interaction and to p = 0.001 for all others

During retrieval, inferior olive activation exhibited an unexpected pattern that appeared to depend on both sequence novelty and retrieval probe type. A post-hoc interaction contrast of sequence novelty (novel vs repeating) and retrieval probe type (match or non-match), with 0 contrast values entered for encoding and maintenance, confirmed that this interaction was significant (F(1, 18) = 4.74, p = 0.043, ηp2 = 0.21). For the repeating condition, non-match probes during retrieval resulted in significantly greater activation than match probes (F(1, 18) = 4.53, p = 0.047, ηp2 = 0.20), as originally predicted. For the novel sequences condition, however, the opposite trend, i.e., match > non-match, was present, although a direct contrast of these means was not significant (F(1, 18) = 2.01, p = 0.173, ηp2 = 0.10). Although retrieval activation for novel matching (M = 2.37, SE = 0.59; t(18) = 4.02, p < 0.001, d = 0.9), novel non-matching (M = 1.20, SE = 0.56; t(18) = 2.13, p = 0.047, d = 0.5), and repeating non-matching (M = 2.62, SE = 0.62; t(18) = 4.26, p < 0.001, d = 0.9) conditions were all significantly greater than 0, the retrieval activation for the repeating matching condition (M = 0.74, SE = 0.56) was not (t(18) = 1.32, p = 0.204, d = 0.3). For the repeating sequence condition, the retrieval matching activation was not significantly different from the maintenance activation (F(1, 18) = 2.03, p = 0.172, ηp2 = 0.10) in contrast to the highly significant maintenance to retrieval increase for the non-matching retrieval phase (F(1, 18) = 11.21, p = 0.004, ηp2 = 0.38). For novel sequences, both matching and non-matching retrieval activations were significantly different from the novel maintenance activations (matching F(1, 18) = 16.38, p = 0.001, ηp2 = 0.48; non-matching F(1, 18) = 10.99, p = 0.004, ηp2 = 0.38).