Distinct neural mechanisms for remembering when an event occurred

Abstract

Events are often remembered as having occurred in a specific order, but almost nothing is known about how the brain encodes this temporal information. It is commonly assumed that temporal information is encoded via a single mechanism, based either on the temporal context in which the event occurred or inferred from the strength of the memory trace itself. By analyzing time-dependent changes in activity patterns, we show that the distinctiveness of contextual representations in the hippocampus and anterior and medial prefrontal cortex was associated with accurate recency memory. In contrast, overall activation in the perirhinal and lateral prefrontal cortices predicted whether an object would be judged more recent, regardless of accuracy. These results demonstrate that temporal information was encoded through at least two complementary neural mechanisms.

Keywords: episodic memory; functional MRI; recency discrimination; temporal context; temporal order memory.

Figure 1

Diagram of experimental task. Thirty-nine object photographs were presented at 6 s intervals in the context of a semantic judgment task. Following a 30 s filled delay, participants were tested with 18 object pairs. Twelve pairs consisted of two objects that had appeared six object apart in the previous list, and participants were to select the object that had appeared earlier. Six pairs consisted of one old and one new object, and participants were to select “New”.

Figure 2

(A) Pattern dissimilarity in the bilateral posterior HPC and right anterior HPC at encoding predicted accurate order judgments at test. (B) Pattern dissimilarity in the medial and anterior PFC interacted with study/test lag to predict accuracy at test. (C) Pattern dissimilarity within the medial PFC predicted accuracy only at shorter lags (< 3 min; t_1,16 = 3.06, p = .004, one-tailed). Error bars denote +/− 1 SEM.

Figure 3

Activation in the right PRC and anterior HPC (A) and left lateral PFC (B) exhibited a significant interaction between Object (Object 1, Object 2) and Accuracy (Correct, Incorrect). (A) In the PRC and HPC, activation was significantly greater for the first object in incorrect pairs (t₍₁₆₎ = 4.02, p = <.001), with no significant difference between the first and second objects in correct pairs (t₍₁₆₎ = 1.02, p = .16). (B) In the lateral PFC, activation was significantly greater for the second object in correct pairs (t₍₁₆₎ = 3.08, p = .004), and marginally greater for the first object in incorrect pairs (t₍₁₆₎ = 1.42, p = .08). Error bars denote +/− 1 SEM.