The psychological reality of frame-shifting
Demonstration of the psychological reality of frame-shifting, then, would suggest a role for pragmatics that goes beyond pragmatics the disambiguator. In particular, frame-shifting suggests that lexical processing does not simply benefit from context, but actively contributes to it. Moreover, background and contextual knowledge do not merely help the listener to specify the meaning of indexicals and disambiguate the meaning of lexical items, but, rather, are crucial for the construction of the message-level representation. Below we discuss the results of studies using three different techniques that establish the psychological reality of frame-shifting: self-paced reading times, eye movement registration, and event-related brain potentials.
Self-paced reading times
To demonstrate the psychological reality of frame-shifting, Coulson and Kutas (1998) conducted a variety of experiments using the self-paced reading time technique. In this experimental paradigm, the task is to read sentences one word at a time, pressing a button to advance to the next word. As each word appears, the preceding word disappears, so that the experimenter gets a record of how long the participant spent reading each word in the sentence.
Stimuli for this experiment were comprised of one-line jokes that required frame-shifting for their comprehension, and straight versions of the same sentences that did not require a frame-shift. Moreover, because we wanted to be able to detect the effect of frame-shifting on the processing of a single word, the disjunctor, or frame-shifting trigger, was always a sentence-final noun. In order to find out what sort of non-joke frames people constructed for these sentences, we performed a norming task in which people were given the jokes minus the last word and asked to complete the sentence with the first word or phrase that came to mind. This is known as a cloze task, and the percentage of people who offer a given word in a given sentence context is known as the cloze probability of that word in that particular sentence context.
Results of the cloze task enabled Coulson and Kutas to ascertain readers’ default (non-joke) interpretation for the sentences. However, it also revealed a disparity in the cloze probability of the most popular response for the items, suggesting that some of the sentence fragments provided a more constraining context than others. For example, (10) elicited a similar response from 81% of the participants, while (11) elicited many different responses, albeit mostly from the gambling frame.
(10) I asked the woman at the party if she remembered me from last year and she said she never forgets a (face 81%).
(11) My husband took the money we were saving to buy a new car and blew it all at the (casino 18%).
As a result, two types of jokes were tested: high constraint jokes like (10) which elicited at least one response with a cloze probability of greater than 40%, and low constraint jokes like (11) which elicited responses with cloze probabilities of less than 40%. To control for the fact that the joke endings are (by definition) unexpected, the straight controls were chosen so that they matched the joke endings for cloze probability, but were consistent with the frame evoked by the context. For example, the straight ending for (10) was name (the joke ending was dress); while the straight ending for (11) was tables (the joke ending was movies). The cloze probability of all four ending types (high and low constraint joke and straight endings) was equal, and ranged from 0% to 5%.
Given the impact of frame-shifting on the interpretation of one-line jokes, one might expect the underlying processes to take time, and, consequently result in increased reading times for jokes that require frame-shifting than “straight” versions of the same sentences. Coulson and Kutas (1998) found that readers spent longer on the joke than the straight endings, and that this difference in reading times was larger and more robust in the high constraint sentences. This finding suggests there was a processing cost associated with frame-shifting reflected in increased reading times for the joke endings, especially in high constraint sentences that allow readers to commit to a particular interpretation of the sentence.