Department of Linguistics

(Joint work with Eytan Zweig, Suzanne Dikker and Hugh Rabagliati) In this talk I discuss a set of magnetoencephalography (MEG) experiments investigating how morphemes are detected from sensory input in the visual modality, the most commonly used modality in studies of syntactic and semantic processing. The first two experiments aimed to identify a potential neural index of pre-lexical decomposition of words into morphemes. Our results indicate that the earliest MEG activity affected by morphological complexity is the right hemisphere generator of the M170 component, associated in previous studies with high-level vision, especially face processing. This morphological complexity effect occurred in the same time-window as the early left anterior negativity (ELAN), an ERP component associated with syntactic category violations and argued to reflect presemantic identification of word category (Friederici, 2002). In the second half of the talk I describe a study that used MEG to elucidate the neural generators of the ELAN. Surprisingly, our results show that violations of word category expectations elicit larger amplitudes already at 100ms in the visual cortex. If already the sensory cortices can check predictions based on syntactic structure, as our findings suggest for the visual modality, this may shed new light on the remarkable rapidity of language processing.

Cognitive theorists acknowledge that mature Theory of Mind involves the understanding of others false beliefs (Dennett, 1978), It is also clear that there are many component processes that contribute to mature Theory of Mind, such as reading intentions and desires, understanding true beliefs vs. pretend, appreciating alternative viewpoints and understanding ignorance versus knowing. Although there is general consensus that coordinating multiple perspectives simultaneously is the fundamental ability required to understand false belief (Perner et al, 2002), neither the target state nor the stages preceding it are well-understood.

Two main kinds of epistemic bias in yes/no (yn)-questions have been discussed in the recent literature. A positive yn-question with the particle 'really', as in (1S'), necessarily conveys an epistemic bias, namely it conveys that the speaker originally believed the correct answer to that question to be in the negative. The corresponding version without 'really' does not necessarily carry this bias and it is thus acceptable in unbaised scenarios like (1), witness the contrast (1S)-(1S') (Romero-Han 2004). Similarly, a negative yn-question with preposed negation necessarily conveys bias towards the positive answer, whereas the version with non-preposed negation does not necessarily carry this bias (Romero-Han 2004). This is shown by the contrast between (2S)-(2S'). Note that this first type of bias presents opposite polarities in the question and in the propositional content of the bias. (1) A(ddressee): Jorge just visited Birgit and Jorn's newborn boy. S(peaker): Did he bring a present for him? S': # Did he really bring a present for him? (Negative epistemic implicature 'He didn't bring a present') (2) Scenario: The speaker is organizing a party and she is in charge of supplying all the nonalcoholic beverages for teetotalers. The speaker is going through a list of people that are invited. She has no previous belief or expectation about their drinking habits. A: Jane and Mary do not drink. S: OK. What about John? Does he not drink (either)? S': # OK. What about John? Doesn't he drink (either)? A second type of bias, discussed in Gunlogson-Büring (2000), is characterized by having the same polarity in the question and in content of the bias. Contextual evidence for p may prompt the speaker to ask the yn-question 'p?' (e.g. 'Is it raining?' in (3S)) rather than 'not p?' or a similar alternative (e.g. 'Is it sunny?' in (3S')): (3) Scenario: Addressee enters Speaker's windowless computer room wearing a dripping wet raincoat. S: What's the weather like out there? Is it raining? (Context evidence bias for 'It is raining') S': # What's the weather like out there? Is it sunny? (Context evidence bias for 'It is sunny') An analysis of the bias in (1)-(2) is developed in Romero-Han (2004), who make crucial use of the conversational epistemic operator VERUM. A second approach has been pursued by Nilsenova-van Rooy (2003) and by van Rooy-Safarova (2003) to cover the data (2)-(3) using Decision Theory. The main goal of this talk is to compare the VERUM-based approach and the Decision Theory approach and to argue that the operator VERUM is needed to derive the data. A second, tentative goal of this talk is to examine the behavior of VERUM / ‘really’ in declaratives and to explain the pattern of answers to biased questions as part of the behavior of VERUM / ‘really’ as an expressive item.

Constraint weighting systems might be extremely promising for phonological analysis. But the process of finding weightings by hand is long, boring, and hard. This is a huge obstacle to fair assessment. We're here to make life easier. We show that constraint weighting grammars reduce to linear programs, which are solvable using computationally efficient optimization algorithms. We present and discuss our implementation of such an algorithm: http://wwwx.oit.umass.edu/~linguist/potts/constraint-weighting You upload two Praat files. The program converts them into a linear system and solves it using the two-phase simplex method. The output is an HTML file displaying the minimal weighting if there is one, else a verdict of 'infeasible'. In the talk, I'll discuss our in-progress extension of the system to typological calculations.