Neural mechanisms of semantic interference and false recognition in short-term memory

Abstract

Decades of research using the Deese-Roediger-McDermott (DRM) paradigm have demonstrated that episodic memory is vulnerable to semantic distortion, and neuroimaging investigations of this phenomenon have shown dissociations between the neural mechanisms subserving true and false retrieval from long-term memory. Recently, false short-term memories have also been demonstrated, with false recognition of items related in meaning to memoranda encoded less than 5s earlier. Semantic interference is also evident in short-term memory, such that correct rejection of related lures is slowed relative to correct rejection of unrelated lures. The present research constitutes the first fMRI investigation of false recognition and semantic interference in short-term memory using a short-term DRM paradigm in which participants retained 4 semantic associates over a short 4-s filled retention interval. Results showed increased activation in the left mid-ventrolateral prefrontal cortex (BA45) associated with semantic interference, and significant correlations between these increases and behavioral measures of interference across subjects. Furthermore, increases in dorsolateral PFC occurred when related lures were correctly rejected versus falsely remembered. Compared with false recognition, true recognition was associated with increases in left fusiform gyrus, a finding consistent with the notion that increased perceptual processing may distinguish true from false recognition over both short and long retention intervals. Findings are discussed in relation to current models of interference resolution in short-term memory, and suggest that false short-term recognition occurs as a consequence of the failure of frontally mediated cognitive control processes which adjudicate semantic familiarity in support of accurate mnemonic retrieval.

Figure 1

Example trials from the ST-DRM task. Positive probes are those that did appear in the memory set. Unrelated lures were not in the memory set, and were unrelated in meaning to items in the memory set. Related lure probes were also not in the memory set, but were semantically associated theme words related to items in the memory set.

Figure 2

Regions showing increased activity for correct rejection of lure vs. unrelated negative probes. As predicted, we found a large cluster of activation in L VLPFC (BA45) associated with semantic interference.

Figure 3

Regions common to both true and false memory are displayed in Panel B. Regions that distinguished true from false memory are displayed in Panel C (red = True Recognition; blue = False Recognition). See Section 3 text for details.

Figure 4

Mean percent signal change (PSC) in our L VLPFC ROI displayed as a function of trial type (A). Individual differences in L VLPFC activity for lure vs. unrelated negative probes was positively correlated with the RT index of semantic interference (B).