Strength of word-specific neural memory traces assessed electrophysiologically

Abstract

Memory traces for words are frequently conceptualized neurobiologically as networks of neurons interconnected via reciprocal links developed through associative learning in the process of language acquisition. Neurophysiological reflection of activation of such memory traces has been reported using the mismatch negativity brain potential (MMN), which demonstrates an enhanced response to meaningful words over meaningless items. This enhancement is believed to be generated by the activation of strongly intraconnected long-term memory circuits for words that can be automatically triggered by spoken linguistic input and that are absent for unfamiliar phonological stimuli. This conceptual framework critically predicts different amounts of activation depending on the strength of the word's lexical representation in the brain. The frequent use of words should lead to more strongly connected representations, whereas less frequent items would be associated with more weakly linked circuits. A word with higher frequency of occurrence in the subject's language should therefore lead to a more pronounced lexical MMN response than its low-frequency counterpart. We tested this prediction by comparing the event-related potentials elicited by low- and high-frequency words in a passive oddball paradigm; physical stimulus contrasts were kept identical. We found that, consistent with our prediction, presenting the high-frequency stimulus led to a significantly more pronounced MMN response relative to the low-frequency one, a finding that is highly similar to previously reported MMN enhancement to words over meaningless pseudowords. Furthermore, activation elicited by the higher-frequency word peaked earlier relative to low-frequency one, suggesting more rapid access to frequently used lexical entries. These results lend further support to the above view on word memory traces as strongly connected assemblies of neurons. The speed and magnitude of their activation appears to be linked to the strength of internal connections in a memory circuit, which is in turn determined by the everyday use of language elements.

Figure 1. Spectrograms of the high- frequency (mir) and low-frequency (mor) words used in the experiments.

The triphonemic consonant-vowel-consonant stimuli were maximally matched acoustically. Note the high similarity between the pitch (solid black line) and intensity (dashed line) contours of the two stimuli.

Figure 2. Event-related potentials elicited by standard and deviant stimuli in the high- and low-frequency deviant conditions.

Midline channels with maximal ERP amplitudes are shown (superimposed on acoustic stimulus waveforms). Note the more pronounced deviant-standard difference in the high-frequency condition, whereas a smaller difference, which was also more focal (here confined to Fz), emerged in the low-frequency condition.

Figure 3. Mismatch negativity (MMN): deviant-standard difference curves for the high- and low-frequency deviant conditions.

Note the more pronounced MMN for the high-frequency word and a smaller MMN for its low-frequency counterpart.

Figure 4. Identity mismatch negativity: direct comparison across experimental conditions between the deviant and standard responses elicited by physically identical stimuli (superimposed on acoustic stimulus waveforms).

Note that the pattern of results (larger deviant-standard divergence, iMMN, for the high-frequency deviant words) is very similar to that in the original MMN analysis (cf. Fig. 2), although here the acoustic differences between the deviant and standard stimuli in each pair have been removed.