Pilot Tests: Determination of Prototypicality and Semantic

were the two pivotal effects in the present experiment, a series of pilot tests were necessitated. Based on two important resource: (A) the categorization proposed by Rosch and Mervis (1975), which includes the six most common categories of concrete nouns in English (determined by word frequency (Kucera & Francis, 1967)), with the prototypicality norms of 50-60 category members for each category being obtained, and (B) the categorization proposed by Snodgrass and Vanderwart (1980), which includes 15 superordinate categories with quite a few exemplars (ordered by

prototypicality). The fifteen superordinate categories are (1) Four-footed Animal; (2) Kitchen Utensil; (3) Article of Furniture; (4) Part of the Human Body; (5) Fruit; (6) Weapon; (7) Carpenter's Tool; (8) Article of Clothing; (9) Part of a Building; (10) Musical Instrument; (11) Bird; (12) Type of Vehicle; (13) Toy; (14) Vegetable, and (15) Insect. Despite the fact that prototypicality of the basic level objects of the 15 superordinate categories is ranked and offered in Snodgrass and Vanderwart (1980), it seems not completely applicable to the present experiment due not only to its lack of semantic relatedness judgment but also to its prototypicality being somewhat

culture-biased. For example, the basic level object “watermelon” is ranked as the least prototypical exemplar in the superordinate category “FRUIT” in Snodgrass et al‟s study while watermelons appear to be very a central/representative one when it comes to fruits, especially for most Asians. This culture-induced mismatch in prototypicality and lack of semantic relatedness triggered the necessity of a series of pilot tests, prior to the real experiment design.

In order to re-rank each exemplar under the superordinate categories in Snodgrass and Vanderwart (1980), a questionnaire about prototypicality, “the Prototypicality Pilot Test” was administered. Thirteen superordinate categories were

47

selected out of the original 15, slightly altered and translated into Chinese. The 13 categories were (1) 動物類 ANIMAL; (2) 餐具廚具類 CUTLERY AND

KITCHEN UTENSIL; (3) 家具類 FURNITURE; (4) 水果類 FRUIT; (5) 武器類 WEAPON; (6) 工具類 TOOL; (7) 衣著類 CLOTHING; (8) 樂器類 MUSICAL INSTRUMENT; (9) 禽鳥類 FOWL AND BIRD; (10) 交通工具類 VEHICLE; (11) 玩具類 TOY; (12) 蔬菜類 VEGETABLE; (13) 昆蟲類 INSECT. The other two categories “Part of the Human Body” and “Part of a Building” were excluded since they were meronomy (holonymy), which was not the focus of this study. All the basic level objects under the selected 13 superordinate categories were preserved⁴ and extra exemplars were added based on the availability of their corresponding pictures. For example, there were 30 basic level objects under the superordinate category “ANIMAL” in Snodgrass and Vanderwart (1980), and 10 extra exemplars with their corresponding pictures available in CRL-UCSD were added to this superordinate category (totally 40 objects) for further prototypicality rating. GOE (Goodness-of-exemplar) ratings were adopted; 0 represented “not a (e.g. “ANIMAL”) at all”, and 5 represented “a very typical (e.g. “ANIMAL”)”. A total of 253 objects under the 13 categories were collected and divided into two questionnaires:

Prototypicality Questionnaire A (126 objects) and Prototypicality Questionnaire B (127 objects) to prevent the raters from carelessly answering due to too many items to finish (See Appendix I-II for the questionnaires). Each object in the questionnaires was rated by 50 students with diverse backgrounds from either Soochow University or National Taiwan Normal University; all the raters were over 18 years old, and they

4 Several cross-categorical objects, such as “CAR”, listed under “WEAPON”, under “TYPE OF VEHICLE”, and under “TOY” were selected only once. As to which superordinate category the basic level object “CAR” should be put under, the rating in Snodgrass et al‟s study was the criteria that was followed (since “CAR” was more likely to be categorized under “VEHICLE”). Therefore, “汽車 (car)”

was placed under the superordinate category of “交通工具類” (VEHICLE) in the Prototypicality Pilot Test.

48

were the college students of different majors, such as English, Japanese, Psychology, Chemistry, Physics, Library Management, or Physical Education, etc. However, all of them were native speakers of Mandarin Chinese. All the objects under the 13

categories were reordered from being the highest (the most central) to being the

lowest (the most peripheral) based on the rating average by the 50 raters. For example, under the category of “FRUIT”, “西瓜 (watermelon: 4.68)” was ranked as the fifth prototypical one (at the topic five), which made a great contrast with the result in Snodgrass et al‟s study.

As for semantic relatedness, another vital role that played in the present study, two pilot tests were conducted. One of them was “Semantic Related Object Naming”, and the other was “Semantic Relatedness Degree Rating”. In the semantic related object naming, what raters had to do was come up with one object which was directly related to the given object. The raters were asked to write down their answers

intuitively without “thinking too much”, and their answers were restricted to nouns only. For example, when a certain object, say “狗 (dog)”, was given, the raters had to name/give an object/noun that was related to “狗 (dog)”, for example, some of the raters gave “貓 (cat)” as a response. All the 253 objects were exactly identical items/objects that were rated in the Prototypicality Pilot Test; in a similar fashion, a total of 253 objects under the 13 categories were divided into two questionnaires:

Related Naming Questionnaire A (126 objects) and Related Naming Questionnaire B (127 objects) to avoid fatigue-triggered imprudence in answering (See Appendix III-IV for the questionnaires). Each object in the questionnaires was rated by another group of 50 students with diverse backgrounds from either Soochow University or National Taiwan Normal University (all the raters were over 18 years old; and the raters were different from the ones in the Prototypicality Pilot Test). The

results/responses were recorded and the frequency of the answer(s) corresponding to a

49

given item was calculated. Due to the answers/responses to the 253 objects rated by the 50 students were quite diverse⁵, the next pilot test for semantic relatedness degree was required.

In semantic relatedness degree rating, the relatedness and unrelatedness among the basic level categories under a certain superordinate category was rated. The

related and unrelated pairs were first selected based on the results of the previous pilot test (i.e. semantic related object naming). Specifically speaking, the top one or two responses to a given object were selected as the (most and second most) semantically related items; what had not been used/mentioned by the raters was chosen to be the most unrelated item. The raters‟ responses often included some objects that were not categorically-related, and these responses were excluded in related/unrelated object pairing. For example, there were 23 different types of responses⁶ to a specific object, say, “烏龜 (turtle)”, with 18 giving “兔子 (rabbit)” as a response, 2 “蝸牛 (snail)”, 5 “龜殼 (turtle shell)”, 2 “鐵鎚 (hammer)”, and so on. Undoubtedly, “兔子 (rabbit)”

was selected as an object that was related to “烏龜 (turtle)” while any other

non-mentioned object, such as “鯊魚 (shark)”, instead of “蝸牛 (snail)”, could be an object that was unrelated to it, since “鯊魚 (shark)” was not from the 23 types of

5 For some prototypical/central objects under “ANIMAL” for example, the answers from the semantic related object naming were more consistent and stable, with 26 raters (out of 50) giving “貓 (cat)” and 6 raters giving “骨頭 (bone)” as a response to the given object “狗 (dog)”, and with 22 giving “狗 (dog)” and 3 giving “老鼠 (rat)” as a response to the given object “貓 (cat)”. However, as for some less prototypical/peripheral objects, like “蝸牛 (snail)”, the answers were even more diverse, sometimes with a case that 50 raters gave 32 different types of responses.

6 Twenty-three different “types” of responses were yielded by 50 raters to the given object “烏龜 (turtle)”. There were 18 people unanimously giving “兔子 (rabbit)” as a response, which was counted as the top one response. Two people gave “蝸牛 (snail)” as a response, which was regarded as the top two response. One person gave “動物 (animal)” as a response, and another one gave “綠蠵龜 (Chelonia mydas) as a response”. The previous four types were regarded as categorically-related responses (i.e. since they were within the same category of “ANIMAL”). The rest of the 19 types of responses are listed as follows, the number in () stands for the number of the raters: 龜殼 (turtle shell);

石頭 (stone) ; 鐵鎚 (hammer); 太陽 (sun); 水池 (pool); 龜速 (turtle‟s speed; as slow as a turtle);

水 (water); 綠色 (green); 水族箱 (aquarium); 速度 (speed); 慢跑 (jogging); 陳銘章導演 (director Ming-Zhang Chen); 海 (sea); 兩棲 (amphibian); 甲骨文 (oracle bone script); 慢 (slow);

比賽 (game/race); 澎湖 (Penghu Islands); 池塘 (pond); these 19 responses were regarded as non-categorically related items, which were excluded in related/unrelated object pairing.

50

responses. Intuitively, “龜殼 (turtle shell)” and “鐵鎚 (hammer)” appeared somewhat related to “烏龜 (turtle)”⁷, but they were out of the selection simply because they were not under the category of “ANIMAL”. Similar to the

Prototypicality Pilot Test, a total of 515 related and unrelated pairs were divided into two questionnaires (See Appendix V-VI for the questionnaires), and then rated based on the GOE rating (1 represented “not very related”, and 5 represented “very related”) by still another group of 50 raters who were also over 18 years old but were

completely different from the ones in the Prototypicality and Semantic Related Object Naming Pilot Tests. A summary of the pilot tests is shown in Table 3.5:

Table 3.5

A Summary of the Pilot Tests Number of the Raters

Number of the Items

Example

Prototypicality 50 253 objects 貓 0 1 2 3 4 5 Semantic Related

Object Naming

50 253 objects 貓 => _____

Semantic Relatedness Degree Rating

50 515 pairs 貓:鼠

0 1 2 3 4 5 貓:大象 0 1 2 3 4 5

To sum up, three experimental tasks were employed in the present study: the Exclusion Task, the Prototype Task, and the Semantic Relatedness Task. Task 1, which used 9 superordinate categories out of the 13 categories, included 24 trials (each coterminous series consisted of 3 basic level objects; a total of 72 basic level

7 A turtle shell “龜殼” is semantically related to a turtle simply because they form a meronomic relation. As to the relatedness between a hammer “鐵鎚 (hammer)” and a turtle, it is probably derived from the idiom in Chinese “烏龜怕鐵鎚 (Turtles are afraid of hammers: Even though turtles are protected with their “hard” shells, they seem to be vulnerable to a hammer attack, which is even fiercer.

51

categories/objects were presented) to test subjects‟ taxonomic hierarchy of categories.

Task 2, which used 10 superordinate categories out of the 13 categories, included 24 sets (each set consisted of 4 basic level objects: one central, one peripheral, and two fillers; a total of 96 basic level categories/objects were presented) to investigate subjects‟ taxonomic hierarchy and prototypicality effects within categories. The last task, which used 9 superordinate categories out of the 13 categories, included 24 sets (each set consisted of 3 basic level objects: one identical, another related, and the other unrelated; a total of 72 basic level categories/objects were presented) to examine relatedness effects within categories. A series of pilot tests for the determination of the taxonomic hierarchy, prototypicality and (un-)/relatedness within categories were conducted to obtain norming data necessary and applicable (in the Chinese context) for the stimulus manipulation.