Given Solt’s conclusions, it would be suggestive to see how moral adjectives fare in this exper-imental paradigm. As we said in the introduction, there are two main reasons for this, which were already summarized in the introduction: first, even though moral adjectives arguably in-volve a judge, it hardly falls under the judge “roles” discussed by Solt—that is, the judge of a moral statement need not, in general, be an experiencer, perceiver or have any particular emo-tion. Secondly, one might suspect that some people have strongly objective intuitions about morality. This is a reason to expect moral adjectives to pattern with the mixed class.
In order to further motivate the first claim, we may consider the acceptability of for/to PPs, which is different for moral adjectivesvis-à-visadjectives in Solt’s EVALclass. According to Bylinina2017, these PPs denote experiencers and fill an argument slot for certain adjectives, such as PPTs. Indeed, PPTs (e.g. tasty,fun) takefor/toPPs, while some moral adjectives do not.
However, this is not visible at first sight, becausefor/toPPs can also be adjuncts, semantically denoting a doxastic operator (something along the lines of in x’s opinionor according to x).
Consider the following sentences:
(5.17) The test was fun for me.
(5.18) The soup was tasty to Alice.
(5.19) Torture is unethical for me.
Contrary to appearances, the PPfor mein (5.19) does not have the same syntactic and semantic role that the PPs in (5.17) and (5.18) (Glanzberg2007).10 Firstly, the choice of preposition in (5.17)-(5.18) is idiosyncratic, but not in (5.19) (Stephenson2007a, p. 520): swappingforforto in (5.17)-(5.18) results in bad sentences, while it is acceptable with (5.19):
(5.20) ?? The test was fun to me.
(5.21) ?? The soup was tasty for Alice.
(5.22) Torture is unethical to me.
Secondly, adjuncts are separable from the main verb by other adjuncts, while arguments are not; an adjunct like yesterday cannot be inserted between a verb and its syntactic argument.
But adjuncts can be inserted in such way:
(5.23) I ate an [arg apple ] [adj yesterday ] (5.24) # I ate [adj yesterday ] [arg an apple ] (5.25) I ate an apple [adj yesterday ]
(5.26) I ate an apple [adj quickly ] [adj yesterday ]
Predicates like tastyand fun display this pattern: (5.27) and (5.28) are fine, but inserting the adjuncts between the adjective and the judge PP makes the sentences bad:
(5.27) The test was fun [j−P P for me] [adjin spite of the baby logic bit ]
10The distinction between arguments and adjuncts isn’t clear cut in general, and even less so with respect to for/to PPs. In particular, the for/toPPs that we observe here are all optional, which is already a feature that syntactic arguments generally lack. We thank Manuel Križ for useful comments.
(5.28) The soup was tasty [j−P P to Alice ] [adj in spite of the celery ] (5.29) # The test was fun [adj in spite of the baby logic bit ] [j−P P for me ] (5.30) # The soup was tasty [adj in spite of the celery ] [j−P P to Alice ] By contrast, both constructions are fine withunethical:
(5.31) Torture is unethical [j−P P for me ] [adj under any circumstance ] (5.32) Torture is unethical [adj under any circumstance ] [j−P P for me ]
We side with Bylinina in suggesting that the behavior of the PPs of (5.17)-(5.18) (idiosyncrasy;
inseparability) points to a thematic relation between those phrases and the relevant predicates.
By contrast, the lack of prepositional idiosyncrasy and the separability data suggests that the for/toPPs that we see in sentences like (5.19) are adjuncts, and in particular, doxastic operators.
What (5.19) says is that, in the speaker’s opinion, torture is unethical. In sum, moral adjectives likeunethicaldo not admit thematicfor/to-PPs.
We take these observations to provide further evidence that moral adjectives lack an experiential semantics. Thus, in this experimental study we set out to test how moral adjectives fare with respect to Solt’s experimental paradigm. To do that, we took a sample of adjectives from Solt’s experiment, added an approximately equal number of moral adjectives, and tested them under Solt’s disagreement paradigm.
5.3.1 Participants
Participants were 40 native speakers of English, recruited via Amazon Mechanical Turk (MTurk). They were paid $0.50 for their participation (the task took approx. 5 minutes).
Recruiting was limited to MTurk workers with U.S. IP addresses. No participant was excluded.
5.3.2 Materials
Test items were based on 24 adjectives. From these, 13 were sampled from Solt’s study:
1. RELNUM: tall, expensive 2. ABS2:empty, full
3. ABS1:salty, wet 4. RELNO:light, hard
5. EVAL:tasty, ugly, happy, intelligent
These were supplemented with 11 moral adjectives, classified along the following axes:
1. THINmoral adjectives:moral, ethical, virtuous
2. THICKmoral adjectives:coward, generous, loyal, honest 3. NORM/VALUE: important, justified, rational, valuable
The difference between thin and thick adjectives has been discussed, so we will not insist on it (see §4.3.3). The class called NORM/VALUEis an heterogeneous class of all-purpose value and normative adjectives: important is usually applied to things that range from subjective to objective value; justified orrational are used indistinctly for moral, practical or epistemic praise; andvaluableis an umbrella evaluative adjective that can be use to evaluate all kinds of things in all kinds of ways.
For each adjective, a disagreement dialogue was created. For the set of adjectives taken from Solt’s study, the same dialogues were used; for the class of moral adjectives, new dialogues were devised. Here is a sample of the new dialogues:
(5.33) a. General McAdam was more coward than General Smith in that particular battle.
b. No, Smith was more coward.
(5.34) a. Ann’s work is more valuable than Jim’s b. I disagree, Jim’s work is more valuable.
(5.35) a. It is more important to follow Mary’s than Bill’s advice.
b. I disagree, Bill’s advice is more important.
Participants were presented with 11 test items and 13 control items, as well as 12 filler di-alogues split between factual (A: Sharks are mammals; B: No they are not!) and subjective disagreements (A: This restaurant has wonderful service, I love it; B: No, it’s awful). See the Appendix for the full list of critical items.
5.3.3 Procedure
Participants were presented with the following set of instructions:
This study is about disagreements between people. Sometimes when two people disagree, only one of them can be right, and the other must be wrong. For example, in this short dialogue, Speaker A and Speaker B can’t both be right, because Rosa can’t have been born in both July and April.
• Speaker A: Rosa was born in July.
• Speaker B: No, Rosa was born in April.
But sometimes when people disagree, there is no right or wrong answer - it’s just a matter of opinion. Here’s an example:
• Speaker A: Susan looks a lot like her sister.
• Speaker B: No, they don’t look alike at all!
In this HIT, you will see a series of short dialogues between two speakers, A and B. Your task is to say whether there is a right or wrong answer, or whether it’s a matter of opinion.
Please answer based on your intuitions; do not think too long about each question.Do not proceed with this experiment if you are not a native English speaker.
Participants were then presented with dialogues like (5.33)-(5.35) above, and told to choose among the following two options:
• What do you think?
1. Only one can be right, the other must be wrong.
2. It’s a matter of opinion.
Answering 1 was classified as a FACT answer; answering 2 was classified as OPINION. At the end, participants were asked for their country, age, biological sex and were given the opportu-nity to comment.
5.3.4 Results
The proportion of FACT choices per Adjective is illustrated in Figure 5.1; the proportion of FACT choices per Adjective Class is illustrated in Figure5.2.
Following Solt’s analyses, we analysed the responses (FACT vs. OPINION) by modelling response-type likelihood using a logit mixed-effect models (Jaeger2008), with the factor Ad-jective Class as fixed effect (8 levels), and random intercept per subject.
The reference level for this omnibus model (aka the baseline in treatment contrast) was the class RELNUM. The z-scores and p-values reported are those calculated by thelme5package by a Wald III test.
The results of this omnibus model indicate that all classes are significantly different from REL -NUM, with the exception of ABS2. Model output is provided in Table5.1.
RelNum (intercept) Abs2 z = −0.62, p=0.53 Abs1 z = −2.82, p<.01 RelNo z = −2.31, p<.05 Eval z = −8.13, p<.001 N/V z = −8.958, p<.001 Thin z = −7.95, p<.001 Thick z = −7.3, p<.001 Table 5.1: Results of Omnibus model
These results roughly replicate Solt’s original findings, modulo the new classes added in our experiment.11
In order to address our main question (i.e., where do moral adjectives fall in the subjectivity spectrum), we fit a second set of models where we compared each of the classes included in Solt (2018) to the new sets of adjectives (THIN, THICK and NORM/VALUE). These models were constructed in the same way as before (fixed and random structure) differing only in how the baseline was encoded.
• ABS1 (reference level) is significantly different from N/V (z = −7.240, p<.001), THICK
(z= −5.3, p<.001) & THIN (z= −6.054, p<.001).
• RELNO(reference level) is significantly different from N/V (z = −7.71, p<.001), THICK
(z= −5.9, p<.001) & THIN (z= −6.42, p<.001).
11We have also reproduced the posthoc pairwise comparisons done by Solt(2018)(ABS1 vsEVAL:z= −610, p<
.001, ABS1 vsRELNO:z=.5, p=.5, EVALvsRELNO:z= −5.8, p<.001)
• EVAL (reference level) is not significantly different from N/V (z = −1.78, p = .07), THICK(z=1.71, p=.08) & THIN (z=.35, p=.072).
Figure 5.1:FACT choices per adjective
Figure 5.2:FACT choices per adjective class