Abstract
Emotive-factive predicates, such as surprise or be happy, are a source of empirical and theoretical puzzles in the literature on embedded questions. Although they embed wh-questions, they seem not to embed whether-questions. They have complex interactions with negative polarity items such as any or even, and they have been argued to preferentially give rise to weakly exhaustive readings with embedded questions (in contrast with most other verbs, which have been argued to give rise to strongly exhaustive readings). We offer an empirical overview of the situation in three experiments collecting acceptability judgments, monotonicity judgments, and truth-value judgments. The results straightforwardly confirm the special selectional properties of emotive-factive predicates. More interestingly, they reveal the existence of strongly exhaustive readings for surprise. The results also suggest that the special properties of emotive-factives cannot be solely explained by their monotonicity profiles, because they were not found to differ from the profiles of other responsive predicates.
Introduction
In formal semantics, "emotive-factive" predicates—verbs that express an emotion toward a fact, such as regret or amaze—have long presented two primary puzzles. First, the Selectional Puzzle: why do these verbs reject whether-questions (e.g., "It surprised Mary whether John came") while accepting wh-questions (e.g., "It surprised Mary who came")? Second, the Exhaustivity Puzzle*: do these verbs trigger "Weakly Exhaustive" readings (being surprised only by the facts that are true) or "Strongly Exhaustive" readings (being surprised also by what is false)?
Historically, these questions were addressed through researcher intuition. This paper employs quantitative psycholinguistic experiments to provide a more rigorous empirical foundation for these linguistic claims, moving beyond "armchair linguistics" to observe how native speakers actually process these structures.
Experimental Methods
The researchers conducted three distinct experiments using participants recruited via Amazon Mechanical Turk. To ensure that world knowledge did not interfere with grammatical judgments, Experiments 1 and 2 utilized a fictional context involving aliens.
Acceptability and Predicate Types
In Experiment 1, participants rated the "naturalness" of various sentences on a continuous slider from "weird" to "natural." This tested the limits of which question types can be embedded under different verb categories. The study systematically compared emotive-factives against other classes such as responsive, anti-rogative, and rogative predicates.
Monotonicity and Truth-Value Judgments
Experiment 2 focused on "Strawson-entailment" to determine if verbs are upward- or downward-entailing. Experiment 3 used a controlled "card game" scenario to detect specific semantic readings. By asking participants to judge the truth of sentences (e.g., "Jacob was surprised by which of his cards are clubs") based on visual evidence of a character's expectations versus reality, the authors could mathematically distinguish between Weakly and Strongly Exhaustive readings.
Key Findings
The experimental results challenge several long-standing assumptions in semantic theory:
- Soft Selectional Constraints: The restriction against whether-questions under emotive-factives is not an absolute grammatical wall. While rated as less natural than know whether, these constructions were significantly more acceptable than baseline ungrammatical controls, suggesting a "degradation" rather than total ungrammaticality.
- Evidence for Strong Exhaustivity: Contrary to the traditional "Weakly Exhaustive only" hypothesis, results for surprise showed a significant presence of Strongly Exhaustive (SE) readings. Participants often judged it true that a character was "surprised by who showed up" if they were surprised by an expected person's absence.
- Monotonicity and Selection are Independent: Both surprise and forget were confirmed as Strawson-downward-entailing. However, because forget allows whether-questions while surprise generally does not, the authors conclude that a verb's entailment profile (monotonicity) is not the secret key to its grammatical selection properties.
Conclusion
This study demonstrates that emotive-factive predicates are more flexible than previously theorized, particularly regarding their ability to trigger strongly exhaustive readings. The findings suggest that the unique grammatical constraints of these verbs cannot be explained by entailment patterns alone. Future research should investigate how these findings relate to Negative Polarity Item (NPI) licensing and explore potential dialectal variations in the acceptability of embedded questions.
Reference: Cremers, A., & Chemla, E. (2017). Experiments on the acceptability and possible readings of questions embedded under emotive-factives. Natural Language Semantics, 25(3), 223-261.
Summary updated