An open question in the pragmatics and psycholinguistics literature is whether asserted and presupposed content differ in the way that they are processed. As a starting place, we'll look at a study that tries to understand the time course of semantic and pragmatic comprehension.
This question of when and how comprehenders works out the meaning of what they hear is part of a much larger question in linguistics and cognition: namely, the question of modularity. Modular accounts in psycholinguistics posit that the systems for interpreting different levels of linguistic structure are encapsulated, use different representations, and operate independently. Under this view, listeners first divide up the sound stream into phonemes, then figure out the word boundaries, then use those words to build up a syntactic structure, then assign a semantic meaning to the sentence, and finally turn to the context to see what pragmatic information may be relevant. Non-modular (a.k.a., integration-based, cue-based, constraint-satisfaction) accounts permit more interaction across levels.
On one hand, there is evidence that supports a view of the mind as highly compartmentalised, with separate systems accounting for processing at the sound, word, syntax, and semantics levels (e.g., patients for whom brain damage yields impairment in one but not other linguistic functions). On the other hand, there are many studies that point to the integration of multiple cues during processing. The study described below (Tiemann et al. 2011) uses the time course of processing as an index of a distinction between semantic processing and pragmatic processing. The study asks whether the pragmatic work that comprehenders do starts during or after the other sentence-internal work. In other words, does semantic processing precede pragmatic processing (as per a more modular account) or are the two intertwined?
Consider an example like (1):
(1) Pete knows that Sue is a linguist.
Under Heim & Kratzer's compositional approach to presupposition two components are posited for the meaning of (1) which can be evaluated for world w:
λw: Sue is a linguist in w.
Pete believes that Sue is a linguist in w.
An alternative (though not incompatible) approach models use rather than formalism: Stalnaker and von Fintel consider the appropriateness of (1) in a particular context. The intuition is that (1) can be added to a context only if the context entails the presupposition (that Sue is a lingiust). In other words, there must be some information (linguistic or non-linguistic) in the context that guarantees that Sue is a linguist. If that is the case, then (1) can be used felicitously.
Under both approaches, the pragmatic processing required to understand the presupposition depends on some syntactic and semantic processing: The meaning of the individual words must be understood (what 'know' means, what 'linguist' means, who 'Pete' is, etc.) as well as the way in which those words combine (e.g., that it is Pete who is doing the knowing, not Sue). If pragmatic processing happens only after the full sentence has been assigned a syntactic structure and the meaning of that structure has been determined, then context-driven effects would not be expected to appear mid-sentence. However, if pragmatic processing happens before the sentence is complete and before all syntactic/semantic work has been finished, there are several key places in the sentence where the comprehenders' work might be apparent.
If pragmatic work is interwoven with the work required to build up sentence-internal structure and meaning, that work is predicted to have the following repercussions. In (1), the word that triggers the presupposition is "knows". Triggers signal to comprehenders that they will need to confirm that the context contains the presupposition. Therefore, words that trigger presuppositions ("know") are predicted to be more difficult than words that require no presupposition ("believe"). The prediction is that non-triggers will be easier to process than triggers, which in turn will be easier to process than semantically unacceptable words.
In (1), it is the word "linguist" that will be treated as the critical word, the point where the comprehender must verify whether the context contains the presupposition. Presupposed content that is true is predicted to be easier to process than presupposed content that is false.
Lastly, if the context contains no information about the truthhood/falsehood of the presupposition, then the comprehender must add the presupposed proposition to the context. Tiemann et al. predicted that the work of adding presupposed content via accommodation is harder than verifying or falsifying the content against the existing context, so verify-context/falsify-context conditions were predicted to be easier than neutral contexts.
Ease will be tested with both offline and online metrics. A rating task will provide an offline measure of how easy it is to accept a particular sentence. Self-paced reading times will provide an online measure of how easy it is to process a sentence. In case you're unfamiliar with this (bizarre) way of reading, self-paced reading experiments ask participants to read sentences one word at a time by clicking on a button to reveal each word; the interval between button clicks is timed and is taken as an index of comprehension ease/difficulty.
The studies were conducted in German using the following triggers:
To go on to section 4.2 on the "Trigger study", click here.