Since its initial discovery by Kutas and Hillyard in the 1980s [1, 2], the N400 component – which is argued to index semantic integration, lexical access, or both (for a thorough discussion, see [3] ) – has been found sensitive to a number of factors including semantic anomaly, cloze predictability, lexical frequency, etc. From a methodological standpoint, one annoying factor that influences N400 amplitude is repetition. That is to say, on subsequent repetitions of a word, the amplitude of the N400 is attenuated (for a review see [4]). This is particularly annoying because repetition interacts with other factors, such as semantic congruity, effectively reducing or suppressing typical effects. This in turn forces researchers to use novel, non-repeated words for each stimulus. Consider a typical 2x2x2 design resulting in 8 conditions. Assuming the bare minimum of 35-40 stimuli per condition, that equates to 280-320 unique words balanced and controlled for the usual suspects. Wouldn’t it be awesome if we could just repeat the same four words time and time again, in lieu of 300 novel words?
A new study by Renoult and Debruille may have found just such a solution. I’ll briefly recap the study here, but if you want to read the original, it’s currently available at JOCN Early Access and is entitled “N400-like potentials and reaction times index semantic relations between highly repeated individual words” [5]. The authors reasoned that semantic congruity effects might be able to resist massive repetition of words if the task required explicit access to semantic representations. In this study, participants were presented visual stimuli consisting of three words: Attention (which initiated each trial), a Category prime word (either
animal or
thing), and a target word (either
dog or
table)
. The Category prime words were presented in lowercase and the target words had a 50% chance to be presented in either upper or lowercase. Participants went through 2 blocks of testing. In the semantic instruction block, participants had to respond by button press to indicate if the target semantically matched or mismatched the category prime word (Fig. 1, A & B). In the physical instruction block, participants had to indicate by button press whether that target word matched the prime word in letter case (Fig. 1, C & D). Each word (animal, thing, dog, table) was presented 200 times in each block.
 |
| Fig. 1 Example of stimuli |
The behavioral data revealed what one would typically expect: (1) when target letter case matched prime letter case, RTs were faster; (2) a trend towards shorter RTs for the physical instruction compared to the semantic instruction; (3) a main effect of semantic match – faster RTs when targets semantically matched their primes; but also (4) a match x instruction interaction, where RTs were faster for the semantically matched targets during the semantic instruction compared to the physical instruction block.
 |
| Fig. 2 semantic match effect modulated by task (adapted from Renoult & Debruille). |
In the N400 window, electrophysiological data revealed that a significant interaction between semantic match and instruction. As illustrated (Fig. 2, top), during the semantic instruction block, semantic mismatches, compared to semantic matches, elicited a negatively deflected N400-like waveform. However, during the physical instruction block there were no differences in amplitude between the semantically matched and mismatched conditions. This is consistent with another study [6] which found that semantic category effects are more reliable when participants are engaged in an explicit semantic task. In order to test whether the match effect was modulated by number of repetitions, the researchers divided their trials into 5 levels of repetition (e.g., 0-20, 20-40, etc.); no interaction of match and repetition level was observed. In contrast to the semantic match effects which modulated the N400-like component, the physical match effects affected the P200 and LPC components, but not the N400. The lack of an N400 semantic match interaction with repetition, suggests that this paradigm (massively repeated prime-targets with an explicit semantic task) was successful in eliminating the repetition effects observed in previous studies. The lack of interaction for semantic match during the physical instruction block and the lack of a physical mismatch effect within the N400 time window are consistent with N400 behavior. ICA analysis revealed generators for semantic match that were similar to those reported for the N400. However, the N400-like effect was shorter in duration and quicker in peak onset than that reported for classic N400 manipulations, which the authors interpret as indicating that the massive repetition of words may render later N400 processes inactive.
In addition to drastically reducing the complexity of experimental design, this protocol may be useful in examining the semantic relationship between individual words as well as between-subjects differences in semantic associations. Before this protocol was developed there was no obvious way for examining the brain’s response to a specific word – the signal from a single word presentation would be lost in EEG noise, and repetition of multiple instances of the word would cause an attenuation of the signature, thus smearing out the “true” response. One can easily imagine a host of similar studies investigating aspects of sociolinguistic cognition. For example, a study investigating cross-dialectal differences in semantic representation: British and American participants see the prime category sport followed by the target football or the prime category thing followed by the target torch. Any differences in ERP waveform would presumably reflect differences in semantic representation (e.g., American – “NFL” and “fire on a log”, British – “soccer” and “flashlight”). The obvious difficulty designing this experiment would be to find words that have the same frequency in both British English and American English. Any other ideas? Anybody want to do this experiment with me? Anyway, a very cool paradigm, and I am sure we will see much more of in the future.
References
1. Kutas, M. and S.A. Hillyard, Reading senseless sentences: Brain potentials reflect semantic incongruity. Science, 1980. 207: p. 203-205.
2. Kutas, M. and S.A. Hillyard, Brain potentials during reading reflect word expectancy and semantic association. Nature, 1984. 307(5947): p. 161-163.
3. Lau, E., C. Phillips, and D. Poeppel, A cortical network for semantics: (de)constructing the N400. Nature Reviews: Neuroscience, 2008. 9: p. 920-933.
4. Rugg, M.D., The effects of semantic priming and word repetition on event-related potentials. Psychophysiology, 1985. 22: p. 642-647.
5. Renoult, L. and J.B. Debruille, N400-like Potentials and Reaction Times Index Semantic Relations between Highly Repeated Individual Words. Journal of Cognitive Neuroscience, 2009: p. 1-18.
6. Devlin, J.T., et al., Anatomic Constraints on Cognitive Theories of Category Specificity. Neuroimage, 2002. 15(3): p. 675-685.
