Here’s a summary of the article “When some is actually all: Scalar inferences in face-threatening contexts” about… scalar implicatures in face-threatening contexts, published 2009 in Cognition.

Instead of following up on the debate over the default character/local computation of scalar implicatures, this paper addresses the question of which kinds of contexts lead to strengthening or weakening of “some but not all” implicatures. Specifically, the authors conduct three sentence rating studies to investigate whether face-threatening contexts (for the listener) lead to a reduction in implicature endorsements. -…they do.

In experiment 1, they show that people are less likely to derive scalar implicatures in face-threatening contexts. In experiment 2, they show that it is indeed threat of face loss, rather than lower-boundedness of the context, that leads to this reduction in implicature derivation. Finally, in experiment 3 they show that listeners are aware of the speaker’s intention (not to “hurt their feelings”) and that they are violating the maxim of quantity (by not deriving the implicature).

Experiment 1: Subjects read two stories – one in the face-threat condition, one in the face-boost condition. Example story:

“Imagine that you have joined a poetry club, which consists of five members in addition to you. Each week, one member writes a poem, and the five other members discuss the poem in the absence of its author. This week, it is your turn to write a poem and to let others discuss it. After the discussion, one fellow member confides to you that `Some people X your poem’.”

In the face-boost condition X was replaced with “loved”, in the face-threat condition with “hated”. Participants then answered the yes/no question: “From what this fellow member told you, do you think it is possible that everyone hated/loved your poem?”
In the face-boost condition (love), 83% of participants said ‘no’ (reflecting the implicature). In contrast, in the face-threat condition (hate), only 58% said ‘no’. They preliminarily conclude that people draw less scalar inferences from “some X-ed” to “not all X-ed” when “X” threatens the face of the listener.

The authors then raise the question of whether the face-threatening contexts in experiment 1 are confounded with lower-boundedness. The argument goes as follows: when evaluating a poem, that everyone loved it is more informative than that some did (because it makes us confident that it really is good). But knowing that everyone hated our poem is not more informative than knowing that some did – in either case we would be downcast enough to burn the poem up on the spot. So, they say, sentences like “some people hated your poem” would evoke a lower-bound context irrespective of any politeness considerations. (I don’t agree with this – why shouldn’t it be the same as in the “love” case – the more people hate my poem, the less confident I become that it is worth anything?)

This confound is addressed in experiment 2 by introducing an additional condition: it is either the listener or a third person who’s face is boosted/threatened by the speaker’s statement. If it was indeed the lower-boundedness of the hate-context that led to the reduced implicature generation, then this should be reflected in experiment 2 in both lower-bounded stories (irrespective of whether the face target is the listener or an other person). If however the effect is due to politeness considerations, reduced implicatures should be observed only in the listener-threat condition, but not in the other three. This is what the authors find, though the methodology is slightly different than in experiment 1: instead of answering a YES/NO question participants rated on a scale from 1 (totally unlikely) to 10 (totally likely) how likely it was that the speaker would use the word ‘some’ if she knew the number of people who loved/hated the poem/chili/organized trip. Participants gave a rating for each number of people (from 1-6). This resulted in fuzzy membership functions for ‘some’ in each of the four conditions – these were almost overlapping in the boost and the other-threat condition, but dropped off faster in the listener-threat condition.

In experiment 3, the assumption was tested that listeners acknowledge the speaker’s intention of saving the listener’s face by using ‘some’ instead of all. Participants rated stories similar to those from experiment 1 on 4 different 10-point scales for how accurate, considerate, honest, and nice it was of the speaker to use ‘some’ if he knew that either 3/6 or 6/6 people loved/hated something the listener had done. Accuracy and honesty ratings were collapsed to a precision index, considerateness and niceness ratings to a kindness index. Precision ratings were very low in both the face-boost and face-threat conditions for 6/6 (in comparison they were very high for 3/6), indicating that people understood that Quantity was being violated. However, kindness ratings were high for 6/6 in the threatening condition. Together, these findings suggest that people are sensitive to both the overall informativeness and the politeness considerations involved in the choice of ‘some’ vs ‘all’.