Sentence Processing

There is some debate in the literature about whether Mandarin topic constructions are movement-derived or base-generated (Huang, Li, & Li, 2009). This study uses the filler-gap paradigm of Traxler & Pickering (1996) to record ERPs in Mandarin filler-gap topic constructions. We use manipulate the plausibility of the filler and the presence or absence of syntactic islands. We predicted an N400 at the critical verb in the non-island conditions, as well as an ELAN/P600 complex at the actual object. Our preliminary results find no N400, but we do see anterior negativity at the verb and a P600 at the object in non-island relative to island conditions. We interpret the P600 to indicate active gap-filling and reanalysis in these constructions.

People generally find it quite easy to construct and understand a negated sentence such as This car is not a Ford. We utter sentences like this many times a day. Yet exactly how negation is processed is a matter of debate among linguists. When we hear a sentence like This car is not a Ford, do we compute the meaning of the structure one word at a time in the exact order the words appear? Or do we compute the meaning of the “inner proposition” first – This car is a Ford – and then, as a second step, negate this proposition?

Using ERPs, we test whether negative sentences are processed in the same way as the positive sentences, or whether processing negation is a two-step process where negation is evaluated after the inner proposition is understood first. 

Previous studies have tested these two competing hypotheses by using negation in a pragmatically licensed environment (Nieuwland and Kuperberg, 2008), or by presenting negated sentences out of the blue (Fischler et al.,1983). In this study, we use a slightly modified version of Fischler et al.’s stimuli embedded in a context. We hope to find an answer to a few questions by evaluating the effect of context on the processing of negation:

• Is negation processed incrementally or non-incrementally (in a two-step fashion)? Does contextually-constrained negation elicit N400?

• Do contextually constrained pseudo-words trigger N400 on a semantically incongruous category word?

We look for an N400 – an ERP response elicited in response to semantic incongruity (Kutas and Hillyard, 1980, 1984; Polich et al. 1981 among several others). By manipulating the truth or falsity of the proposition, we can determine whether the brain computes the meaning of the inner proposition first, or whether it computers the meaning of the words in the order they appear. 

The final word of the sentence will either be congruous or incongruous, depending on whether it is negated. A robin is a bird is perfectly congruous, but A robin is not a bird is false and unexpected! By contrast, A hammer is a bird is incongruous, but A hammer is not a bird is true and perfectly congruous. The brain’s response to this final word will tell us whether negation is computed before or after this word is integrated into the meaning of the sentence.

If negation is computed before reaching the final word of the sentence, only the false sentences (regardless of negation) will elicit an N400. If negation is computed after the meaning of the inner proposition has been computed, then the N400 is expected in false-affirmative (A robin is a tree) and true-negative sentences (A robin is not a tree).

We test two groups of participants. One group will read sentences using real words (robin, hammer). The other group will read complementary versions of these sentences with pseudo-words (dwopse, twizen). We are interested in whether context will affect the order in which negation is computed, and whether this process will be the same for words that have to be learned in real time.

Our preliminary results support non-incremental processing of negation in the pseudo-word group (N=13). This also indicates that the semantic mapping of a newly learned word in a context is very fast – the newly learned words can elicit an N400 immediately after initial exposure.

There has been a debate on the time course of the processing of binding theory and non-linguistic cues. Does binding theory act as an initial filter during processing, followed by the second stage of processing where ungrammatical antecedents can interfere with the initially established binding relation (Sturt, 2003)? Or, are all possible antecedents activated during initial processing, with multiple constraint satisfaction mechanisms selecting the ultimate winner (Badecker & Straub, 2002; Runner & Head, 2014)? 

This study expands the ERP study on the processing of anaphor (Xiang et al, 2009), and improves upon that study with fully crossed and appropriately controlled 2x2 designs. Furthermore, we manipulate the non-grammatical, gender cues of grammatical and ungrammatical antecedents in several other ways.