• No results found

2.7 Final conclusions

3.1.2 Procedure

The following experimental procedure was proposed. We conducted two separate sub-experiments checking the above:

• Test 1: so-called existential import of the quantifiers “most” and “all” (pos- itive and negative premises) together with scalar implicatures of “some” – the strong and the weak.

• Test 2: scalar implicatures of “most” (the particular and the universal) to- gether with inferences All/M ostand N o/M ostnot (reversed implicatures). For both experiments we planned two independent groups, which may be treated as “basic” and “control”. Thus there were4tests in total. All four tests consisted (each) of 60 direct inferences to evaluate. Subjects were asked if on the basis of a given sentence they can utter some other sentence:

You know that most A’s are B.

Can you say on the basis of the above sentence thatsome A’s are B? YES / NO

A,B represent any terms used in questions. There was no reference to truth or correctness. Thus the questions checked behavioral disposition of subjects to utter or not given sentences if some other facts are known to them. There were40

tested inferences in each test and20fillers. In all four tests we used as fillers infer- ences similar to the checked ones but logically obviously incorrect. By obviously incorrect we understand classically incorrect inferences, for which one cannot find any plausible explanation for correctness; they are not justified in terms of impli- cature or in terms of any existential load (or the lack of such load) of quantifiers. One example of such obviously incorrect inference is M ost/All. It is hard to think of any model of reasoning in which such an inference would be regarded as correct. Few doubts may arise considering such a choice of experimental layout. The first is that there are only 20 fillers in the set of 60 tasks. However if we regard20plus 20 tested inferences, then it turns out that they work as fillers for each other. The other thing is that we do not use any real fillers – so tasks of a completely different form, that could distract subjects’ attention. The reason is that first of all we did not want to make the tests too long, secondly that we had to instruct our subjects about what we expect from them, namely they were told that they should follow their intuition (language understanding, etc) while solving the tests. We consider that mixing our tasks with tasks of a completely different form would causetoo much confusion and would make the experimental situation artificial.

Test 1 – so-called “existential import” & implicatures of “some”

Test 1 consisted of the following groups:

Test 1a – the “most” group checking the so-called existential import of “most” and the strong scalar implicature of “some”.

Layout:

• M ost/Some (10) and M ostnot/Somenot (10)

• Some/M ostnot (10) and Somenot/M ost (10).

• Fillers:

M ost/M ostnot(5) and M ostnot/M ost(5)

Some/Somenot (5) andSomenot/Some (5)

Test 1b – the “all” group checking the so-called existential import of the uni- versal quantifier, together with the weak implicature of “some”.

3.1. Layout of the experiment 71

• All/Some (10) andN o/Somenot (10)

• Some/Somenot (10) and Somenot/Some (10)

• Fillers:

Some/All (5) and Somenot/N o (5)

All/N o (5) and N o/All (5) We planed the following comparisons:

• between-subject:

M ost/Some with All/Some and M ostnot/Somenot with

N o/Somenot.

Some/Somenot with Some/M ostnot and Somenot/Some with

Somenot/M ost.

• within-subject: with respect to the occurrence of negation, so : M ost/Some

with M ostnot/Somenot etc.

Test 2 – scalar implicature of “most”

Each of two groups checked a different form of a scalar implicature of “most”:

Test 2a – the particular scalar implicature: M ost/Somenot (10) and

M ostnot/Some(10)

Test 2b – the universal implicature: M ost/notAll (10) and M ostnot/notN o

(10).

Both groups – a and b checked additionally the same reversed implicatures and use the same fillers:

• Reversed implicatures: All/M ost (10) and N o/M ostnot (10)

• Fillers

M ost/All (5) and M ostnot/N o (5)

All/M ostnot (5) andN o/M ost (5) We planed the following comparisons:

• between-subject comparisons of the particular and the universal implica- ture.

• within-subject comparisons of scalar implicatures and reversed implicatures, positive and negative forms.

Comparisons between tests

Furthermore, we planed some comparisons between tests (thank to the unified set- up of materials, each of the four groups can be compared with the remaining three, provided that the number of subjects in each group is similar). The following comparisons were the most important:

1. Reversed implicatures with existential import of “all” (resp.“no”) and with existential import of “most” (“mostnot”)

2. Comparison of scalar implicature of “some” and “most”

“Yes” and “No” answers

Note that all tests consist of YES/NO tasks. The first version of the experiment assumed the possibility of giving a “don’t know” answer. However, on the basis of a pilot research (c.a. 5 tests given to random people to solve), we observed that subjects did not tend to give this kind of reply. They chose only “yes” or “no” responses. Therefore we decided to exclude the “don’t know” possibility for the sake of simplifying tests and future statistics. (If it happened that a very small percentage of subjects tended to give “don’t know” replies, then it could be problematic for our statistical analysis.)

Choice of terms; materials

Based on the results from the previous experiment, which proved no significant difference in subjects’ responses with respect to domains (empty versus non- empty terms), we decided to use in the second experiment only empty terms of first category, thus new-introduced, non-existing in language, terms. The main reason for our decision was to minimize contextual factor in the reasoning, so that people would not give answers based on their knowledge about world.

The following 10 pairs (an abstract term plus a property) were chosen and combined with different quantifiers in positive and negative sentences: mer- mogliny + pink,buzaki + green, mroczniaki + bring bad luck,grombliny + have claws, mgłowce + have a cap, zarkotki + have long ears, trakloki + intelligent,

wyszczyki + two-coloured, klawuchy + have red tails, leprokraki + like cheese. Each pair was used for each syntactic form only once, so that no identical task was repeated. In the case of tasks with only 5 trials (fillers), we chose 5 pairs of a term and a property and the remaining 5 pairs were used for another filler. The tasks in each of four types of tests were ordered randomly by a computer program. The English translations of these tests can be found Appendix C.

3.2. Results 73

Groups

Each of the four tests was solved by an independent group of people. These were students aged between 19 and 39 (however only 6 people in total were over 30) from different, but mainly humanistic, faculties. The selection of people assumed that subjects should not have any logical experience – so not only mathematical or logical backgrounds, but also philosophical backgrounds were excluded. Psy- chology students were also not preferable. Subjects in each group were mixed and came from the following backgrounds: history, English philology, Russian philology, pedagogy, culture studies, criminal investigation techniques, law.

Table 3.1 compares group sizes and subjects’ age.

GROUP 1a 1b 2a 2b

number of subjects 49 48 40 40

age average 22.63 21.16 23.5 22.67

age range 19−32 19−32 20−39 20−33

Table 3.1: Groups in the passive competence experiment

3.2

Results

3.2.1

General analysis of results