In this seminar I will discuss MEG and behavioral studies on lexical and morphological processing conducted by myself and/or my colleagues at the KIT/MIT MEG Laboratory at MIT, the CNL laboratory at the University of Maryland and the Linguistics Department in New York University. The continuous theme of the lectures will the use of the MEG response component M350 as a linguistically relevant dependent measure in the study of lexical and morphological processing and representation. The lectures connect to each other very tightly, and are not meant to be independent.
Day 1
Lexical access 1: The M350 as an MEG index of lexical activation
SLIDES.
I will start the seminar by discussing general questions having to do with methodology in cognitive neuroscience and by providing some historical background to electrophysiological research on lexical processing, with focus on the N400. I will then present a series of MEG experiments where lexical factors were manipulated. Our results collectively suggest that the first MEG component affected by lexical factors, but not by post-lexical factors, is the M350, a primarily left-lateralized response component peaking at 300-400ms post word onset. I will argue that the M350 indexes the initial activation of the mental lexicon, and discuss the relationship of this claim to electrophysiological evidence from other labs, as well as to evidence from other techniques, such as eye-tracking and masked priming.
Day 2
Lexical access 2: The M350 and mechanisms of recognition
SLIDES.
Behavioral studies on word recognition have unambiguously shown that lexical access involves large-scale activation of lexical representations, most of which are incorrect matches to the input. How does the language processor select the best match to the linguistic input, in a matter of a few hundred milliseconds? In the first half of this lecture I review electrophysiological evidence pertaining to the processing stages of lexical access, including early perceptual processes. In the second half, I show data from a cross-modal priming experiment where we used the M350 to distinguish between different hypotheses about the basic mechanism of word recognition.
Day 3
Morphology 1: The M350 as a tool for investigating similarity and identity
SLIDES.
What constitutes the 'same' or 'different' in human cognition is a fundamental question for the study of the mind/brain. Symbolic approaches to language assume a clearly defined notion of sameness: linguistic computation involves manipulation of a finite set of discrete symbols and therefore 'the cat sat' and 'a fat cat' literally involve accessing the same symbolic representation for 'cat'. In contrast, (at least some) connectionist architectures do not implement a notion of sameness that is distinct from extreme similarity. In this lecture I discuss behavioral and MEG evidence pertaining to the question of whether morphological relatives, such as ‘teacher’ and ‘teach’, interact with each other in processing in ways that suggest that they contain an identical element ‘teach’, as dictated by a decompositional view of morphology, or whether their relationship resembles similarity at the extreme. Both behavioral and M350 data suggest that morphologically related items stand in a fundamentally different relation to each other than lexical representations that are simply similar to each other, lending support to decomposition. I conclude by discussing recent experiments where the M350 was used to tap onto representational identity in irregular verb morphology.
Day 4:
Morphology 2: Electrophysiological and behavioral evidence for early effects of morphology
SLIDES.
If the perception of morphemes, rather than “words”, is the basic mode of language processing, then the relevant frequency measure for lexical activation should be morpheme frequency, not word frequency. Discussion in this lecture will be centered around an MEG experiment where we tested for the presence of cumulative morpheme frequency effects in derivation, and surprisingly found that any facilitory effect of high cumulative morpheme frequency is entirely overshadowed by an early competition effect. Our data suggest that there is competition among morphological relatives, but that this competition occurs earlier than competition among phonological neighbors. I will relate these findings to results from masked priming and ERPs, which also suggest that (at least some) morphological processing occurs very early.