Literal Lucy: Experimental work on scalar implicature

The Truth Conditional Project: Experimental Pragmatics

~ Research Team ~

Rachel Baker (Linguistics, Northwestern (NU)), Matt Berends (Linguistics, NU), Alex Djalali (Linguistics, NU), Ryan Doran (Philosophy, NU), Meredith Larson (Linguistics, NU), Yaron McNabb (Linguistics, U Chicago), and Gregory Ward (Linguistics, NU)

~ Papers ~

Larson, Meredith; Ryan Doran, Yaron McNabb, Rachel Baker, Matthew Berends, Alex Djalali, & Gregory Ward. (in press). Distinguishing the Said from the Implicated Using a Novel Experimental Paradigm. Proceedings for Experimental Pragmatics 2007.

~ Recent Presentations ~

2008 "Distinguishing WHAT IS SAID from WHAT IS IMPLICATED within a Gricean Framework." An invited talk presented by Ryan Doran & Gregory Ward to PhLing, Evanston IL, USA.

2008 "The Effects of Scale Type and Salience on the Interpretation of Scalar Implicature." Presented by Matt Berends & Rachel Baker at the Linguistic Society of America Annual Meeting in Chicago IL, USA. abstract

2007 "Distinguishing the SAID from the IMPLICATED Using a Novel Experimental Paradigm." Presented by Ryan Doran & Yaron McNabb at Experimental Pragmatics 2007 in Berlin, Germany. abstract

2007 "Distinguishing among Contextually-Determined Aspects of Utterance Meaning: An Empirical Investigation ." Presented by Ryan Doran & Meredith Larson at the Linguistic Society of America Annual Meeting in Anaheim CA, USA. abstract

~ Background ~

This project grew out of a seminar class on GENERALIZED CONVERSATIONAL IMPLICATURES (GCIs) during the 2006 Winter Quarter. Our primary interest was to determine whether the categories theorists presented in the literature were psychologically real. Specifically, we wanted to know how non-specialists interpret various purported implicatures (e.g. whether they treated them like entailments, whether they categorized them according to rubrics proposed in the literature, such as Horn (1984) or Levinson (2000)). Furthermore, we wanted to see if we could improve on some of the previous experimental methodologies. To this end, we created a new experimental paradigm, which we call "Literal Lucy".

~ Example of the paradigm ~

In our paradigm, participants are introduced to the character Literal Lucy, a literal-minded person who has a tendency to interpret figurative language and indirect speech acts literally, as in (1) below:

                        (1)
                        Frank: Hey, Lucy, can you tell me what time it is?
                        Literal Lucy: Yes, I can!
                        Frank: So...?

In our experiments, participants read various short conversations between two fictional characters, Irene and Sam. Participants then evaluate the truth of underlined sentences in the stories from the perspective of Literal Lucy. During the actual experiment, participants would see conversations such as (2) below and were instructed to evaluate the underlined sentences' truth from the perspective of Literal Lucy*. Participants then rated their confidence in their answers.

                       (2)
                        Irene:    Hey, Sam. Do you know who wrote Pride and Prejudice?
                        Sam:      A British woman wrote it, and her last name was Austen.
                        FACT:    Jane Austen, a British woman, wrote Pride and Prejudice.

*There were two experimental conditions in the first experiment that did not include Lucy, as will be explained below.

~ Experiments ~

Experiment 1 - Investigating variety in pragmatically-sensitive meaning with a new paradigm
To see examples of our materials, see our handout.
We designed this experiment with the following three main goals: (i) to determine whether speakers can isolate a level of meaning corresponding to the Gricean notion of what is said that is separate from the contributions of GCIs; (ii) to look for empirical evidence for the various distinctions among GCI types found in the literature; (iii) to improve upon previous methodologies by developing a paradigm that can train speakers to distinguish the truth-evaluable meaning of an utterance from implicatures without relying on theoretical terms such as what is said.

To address (i), we used a truth-evaluation task instead of a "meaning"-interpretation task. This task and the confidence rating were designed to determine whether participants were able to distinguish among different types of meaning, e.g. entailments versus implicatures.

To explore (ii), we used multiple types of pragmatically-sensitive meaning, including both Necessary Contextual Elements (NCEs), which included ellipsis, pronoun resolution, indexicals and deictics, and GCIs, which were based on examples from the literature and included categories such as quantifiers and modals, cardinals, gradable adjectives, verbal periphrasis, bridging inferences, conjunction buttressing, for a total of 11 different types. We also used positive and negative entailments as control items.

To achieve (iii), we created three instruction conditions for the True-False task: (a) Literal Lucy, (b) Literal/No-Lucy, and (c) No-Literal/No-Lucy. In the Literal Lucy condition, participants were trained as explained above. In the Literal/No-Lucy condition, participants were not trained with the Lucy character and were instructed to evaluate the truth of the underlined statement interpreted literally. In the No-Literal/No-Lucy condition, participants were instructed to evaluate the truth of the underlined statement without any mention of literalness or Lucy.

Experiment 2 - Investigating the effects of scale types and scale salience
In this experiment, we look more closely at scalar implicatures using the Literal Lucy paradigm. We systematically varied the type of scale and the salience of alternative scale members to see whether such features affected participants' judgments. Scales were classified according to whether they have fixed end points or not (closed/open), and whether they admit intermediate values or not (discrete/continuous), resulting in four scale types: closed continuous (<all, many, some>); closed discrete (<succeed, attempt>); open continuous (<…sweltering, hot, warm…>); and open discrete (<…three, two, one>). In addition, we varied how the scale was evoked, using a 3-way distinction, leading to a 2x2x3 design.

Contact us ~
If you are having problems with the website or links, email Meredith at meredithjlarson [@] gmail.com

This Web Page Created with PageBreeze Free Website Builder