國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
51
CHAPTER 4 RESULTS
The results are organized in three sections. First, in 4.1, there is the result of sense analysis, which shows the sense distribution of in and on used by native speakers and learners. The frequencies of the senses appearing in the data were calculated to examine how in and on are used by these two groups. Then, in the result of semantic feature analysis, presented in 4.2, there are descriptive data for the
semantic features (e.g. concrete, animate) in the retrieved instances. The frequencies of semantic features in the nouns contained in the prepositional construction of in and on by native speakers and learners were computed and analyzed by employing
descriptive and inferential statistics. Lastly, we also examine learners‘ error in constructing in and on prepositional phrases in 4.3.
4.1 Results of the Sense Analysis
Figure 4.1 illustrates the two corpora (BNC and NCCU) that were included in the analysis, and each can be divided into the data of in and on. Within each of these two prepositions, the data can be further categorized into literal or metaphorical
constructions, in which their senses were identified. Within each set of literal or metaphorical data, figure and ground noun phrases will be identified in order to conduct the semantic feature analysis. This section will present the results of the sense analysis, and the next section will show the results of semantic feature analysis.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
52
Figure 4.1 The Layout of the Datasets in the Analysis
The descriptive statistics of the overall data is shown in Table 4.1, which presents raw frequency and percentage of literal and metaphorical expressions of in and on, e.g. a brick would float on water (literal), Myth and menace merge in India‟s poll campaign (metaphorical), in BNC and NCCU. As mentioned in the previous chapter, literal and metaphorical expressions were treated separately since semantic extensions of them would be at different levels.
Table 4.1 Raw Frequency and Percentage of Literal and Metaphorical in and on
Literal % Metaphorical %
IN BNC 446 44.60% 554 55.40%
NCCU 515 51.10% 485 48.90%
ON
BNC 269 26.90% 731 73.10%
NCCU 392 39.10% 608 60.90%
Total 1617 40.43% 2383 59.58%
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
53
In Table 4.1, the overall percentage shows that more metaphorical expressions are used (59.58%) in the data than the literal ones (40.43%). Generally, in BNC, more metaphorical expressions are used, and the metaphorical expression with on accounts for much higher percentages in both BNC (73.1%) and NCCU (60.9%). This
conforms to the fact that when inspecting the senses of in and on listed on
Merriam-Webster Online Dictionary, more non-literal senses may be found with both prepositions. As noted in previous studies on the senses of preposition (as in Evans and Tyler, 2004) that the metaphorical senses are extensions of the literal sense, more metaphorical extensions indicate the semantic complexity of prepositions. However, for metaphorical phrases of in, higher percentage was only be found in BNC (55.4%), while in NCCU, literal phrases accounted for higher percentage (51.1%) although the differences between literal and metaphorical for in are smaller than those of on. With regard to the contrast between in and on, more metaphorical phrases were identified in the data of on than in in. In the following step, sense analysis was then conducted to examine how these literal and metaphorical senses are actually used by language users.
To investigate the relations among the literal and metaphorical senses, the current researcher grouped these senses according to the clusters in the semantic networks constructed by Evans and Tyler (the network of in) and Ho (2007) (the network of on). For senses within each cluster, the hypothesized prominent features of figure and ground nouns are illustrated in Table 4.2, with examples given in the last column. From this table, we can see that slight difference may be found within the hypothesized prominent features of figure and ground in senses of different clusters (see p.13 and p.15 of this thesis for the network of clusters for in and on).
‧
Table 4.2 Hypothesized Prominent Features of ‗Figure‘ and ‗Ground‘ in Prepositional Phrase of in and on3
Prepositional phrase of in
Cluster Prominent Feature
Example
concrete Mrs. Brown lived in Morrell House.
Location
concrete, animate, human, mobile, solid
measure, activity The circuit cable runs direct to each ceiling rose in turn.
Vantage Point is Interior
mobile, solid partitive, countable I have him in sight.
Segmentation
The walls of the sandcastle fell in.
Prepositional phrase of on
Cluster Prominent Feature
Example
Many passengers were on board.
State mobile, activity mobile, activity, temporal
The DVD is on pause.
Support concrete, animate, human, mobile,
concrete, partitive, countable
The movie is based on the true story.
Continuation concrete, animate, human, mobile,
activity, mobile Please don‟t stop, keep on talking.
Within each construction of prepositional expressions in different clusters, there
3 These clusters are from the semantic networks proposed in Evans and Tyler (2004) and Ho (2007), and the features hypothesized here are from Table 3.2, the set of features constructed for the semantic feature analysis.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
55
are figure and ground nouns which can be described through their prominent semantic features. For example, within the PROTO-SCENE cluster of in, the expressions tend to be constructed by animate and human entity as the figure, thus, hypothesized as concrete, animate, human, and mobile features. As for their grounds, the expressions usually involve a concrete container or location (e.g. city, box); therefore, it is
hypothesized to involve mostly concrete feature. With regard to the relative size of figure and ground, as figure tends to situate inside the boundary of ground (e.g. the boy (F) in the house (G), thus, F<G), so it is hypothesized to be smaller than grounds.
Through examining the semantic feature of nouns within senses of these clusters, we can perceive how the nouns constructing the prepositional phrase might influence the semantics of prepositions. These hypothesized prominent features will later be examined in the results of semantic feature analysis to inspect if these listed features are also prominent in the native speaker data.
On the other hand, in order to probe into learners‘ performance on using
preposition in and on, an additional sense analyses based on the data from NCCU and BNC were run, in which the latter were used as the reference data set4. The result of the sense analyses may be compared and contrasted with the results of semantic feature analysis, which will be presented in the next section, for examining whether nouns with particular sensory-motor features may have prominent tendency to appear in the figure or ground nouns in constructing prepositional phrases of particular sense.
This inspection into the dictionary meaning was carried out as an initial analysis of the corpora as the entries in the dictionary are what language users may most
frequently refer to when encountering unknown meanings or usages. This full list of
4 ‗Reference data‘ refers to the data from the reference corpus which is large enough to represent a comprehensive information about a language and is used as a reference for comparison with other corpus (cf. Baker, 2006; McEnery, Xiao, & Tono, 2006).
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
56
dictionary explanations toward a particular lexicon may also facilitate the researcher to examine the multiple facets of a lexical item. However, they may not be
comprehensive in showing the meanings of the preposition, so a corpus-based analysis may complement the explanations listed on the dictionary. To further examine how the list of lexical entries can be related to the semantic network constructed in the previous studies (Evans and Tyler, 2004; Ho, 2007), the current researcher attempted to identify the senses according to the semantic clusters
proposed in the network graphed in Evans and Tyler‘s (presented again in Figure 4.2) and Ho‘s work5.
Figure 4.2 The Partial Semantic Network of in (Evans and Tyler, 2004, p.173)
Table 4.3 below shows the distribution of the senses of in and how the senses can be categorized into Evans and Tyler‘s semantic clusters in the lexical network of in.
5 Ho‘s (2007) network will appear again in the discussion of on on p.57.
‧
Table 4.3 Sense Distribution of in
Cluster Sense BNC NCCU
Literal Metaphorical Literal Metaphorical Proto-
to or toward the inside especially of a house or other building
4% (36) 0% (4) 1% (5) 0% (2)
7
to or toward some destination or a
particular place 0% (1) 0% (3) 0% (0) 1% (6)
8 at close quarters : near 0% (0) 0% (2) 0% (0) 0% (2)
Reflexivity 9
so as to incorporate —often used in combination
0% (0) 1% (5) 0% (0) 0% (2)
Segmentat
ion 10 to or at an appropriate place 1% (5) 0% (4) 0% (0) 0% (3)
Location 11
within a particular place; especially: within the customary place of residence or business
0% (0) 0% (0) 0% (0) 0% (0)
N/A 12
in the position of participant, insider, or
officeholder —often used with on 0% (1) 2% (19) 0% (3) 1% (14)
N/A 13 on good terms 0% (0) 0% (0) 00% (0) 0% (0)
N/A 14 in a specified relation 0% (0) 0% (0) 0% (0) 0% (0)
N/A 15 in a position of assured or definitive success
from a condition of indistinguishability to
one of clarity 0% (0) 0% (0) 0% (0) 0% (0)
Total 446 554 515 485
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
58
The left column shows the corresponding cluster matched to each sense
identified by the current researcher, though some of the senses cannot be identified in these clusters (shown as not available ‗N/A‘). This implies that the semantic clusters may not be sufficient to include and describe all the meanings in the dictionary entry, and not all the semantic clusters have their corresponding meanings listed on the dictionary. Therefore, in the following semantic feature analysis with results
presented in next section, we will focus both on the senses identified in the semantic clusters as well as those that do not have the corresponding clusters, which will be discussed according to the prominent features in figure and ground nouns.
The result of sense analysis for in, as displayed in Table 4.3 above, shows the distribution of the senses analyzed by this work based on the meanings given in the Merriam-Webster‘s Online Dictionary (http://www.merriam-webster.com/), and these counts were separated into literal and metaphorical expressions. Through analyzing large amount of data from the corpora, it may help us detect possible problems with the dictionary meanings. As can be observed in Table 4.3, in both literal and
metaphorical data, most of the instances fall into the category ―inclusion, location, or position within limits‖ in both corpora. However, in Evans and Tyler‘s semantic network, the ―inclusion‖ sense resembles the proto-scene, which initially involves merely the literal meaning of in. For the data categorized under the ―inclusion‖ sense in Table 4.3, they originally refer to both literal and metaphorical meanings without distinguishing these two levels of meanings. This way of categorization may simplify the ―inclusion‖ sense and may pose problems in learning.
The second-most frequently appearing metaphorical sense is Sense 3 ―used as a function word to indicate limitation, qualification, or circumstance‖, and it is two times higher in BNC than in NCCU. However, in other categories, only some slight
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
59
variations can be found between these two corpora. This may imply that native
speakers and learners‘ use of the senses and usages of preposition in does not differ so greatly in the other senses.
For the data of on, the distribution of the senses of on and how the senses can be categorized into Ho‘s (2007) semantic clusters in the lexical network of on (Figure 4.3 below) are shown on Table 4.4.
Figure 4.3 The Partial Semantic Network of on (Ho, 2007, p.69)
Ho‘s semantic network is adopted as a reference for the study of on in this study because her network basically follows Evans and Tyler‘s (2004) network that
categorizes distinct senses according to their semantic clusters, and this network has also been evaluated by Evans (through personal communication, p.86 in Ho‘s work) as a reasonable categorization for the distinct senses of on.
‧
Table 4.4 Sense Distribution of on
Cluster Sense BNC NCCU
Literal Metaphorical Literal Metaphorical
Proto-scene
1 position in contact with and supported by the top surface of
a time frame during which something takes place or an instant, action, or occurrence when something begins or is done
0% (1) 2% (17) 1% (4) 2% (24)
Support
10
manner of doing something; often used with the
(1) active involvement in a condition or status, (2) regularly using or showing the effects of using
0% (3) 5% (50) 0% (0) 1% (7)
14 involvement or participation 0% (3) 6% (59) 0% (0) 2% (15)
15 inclusion 0% (0) 0% (4) 0% (2) 0% (1)
16
position or status in proper relationship with a
standard or objective 0% (2) 1% (5) 1% (4) 0% (9)
Support
17
reason, ground, or basis (as for an action,
opinion, or computation) 0% (4) 8% (81) 0% (2) 7% (67)
18 cause or source 0% (2) 6% (56) 0% (0) 1% (6)
N/A
19 the focus of obligation or responsibility 0% (1) 0% (3) 0% (0) 0% (1)
20
the object of collision, opposition, or hostile
action 0% (4) 1% (11) 0% (1) 3% (32)
21
the object with respect to some disadvantage,
handicap, or detriment 1% (7) 3% (25) 1% (7) 3% (33)
22
destination or the focus of some action, movement, or directed effort
2% (19) 22% (216) 1% (8) 24% (237)
23 the focus of feelings, determination, or will 0% (2) 1% (5) 0% (1) 2% (19)
‧
24 the object with respect to some misfortune ordisadvantageous event
0% (0) 0% (0) 0% (0) 0% (1)
25
the subject of study, discussion, or
consideration 0% (2) 5% (48) 0% (1) 2% (20)
26 with respect to 0% (2) 2% (19) 0% (0) 2% (15)
Continuation 27 reduplication or succession in a series 0% (0) 0% (1) 0% (0) 0% (1)
Proto-scene 28
in or into a position of contact with an upper surface especially so as to be positioned for use or operation
0% (3) 1% (5) 0% (4) 0% (2)
N/A 29
in or into a position of being attached to or
covering a surface 0% (0) 1% (1) 0% (3) 0% (3)
Continuation 30
forward or at a more advanced point in space
or time 0% (0) 0% (5) 0% (0) 1% (5)
31 in continuance or succession 0% (0) 1% (4) 0% (0) 2% (15)
State 32
into operation or a position permitting operation
0% (0) 0% (1) 0% (0) 0% (3)
Total 269 731 392 608
As the table for preposition in, the left column in Table 4.4 displays the
corresponding cluster of each sense identified by the current researcher. In the central sense ―in contact with or supported by a surface‖ (Sense 1), NCCU accounts for higher percentage in literal phrases (4%) than BNC (2%). Similar result can also be observed in senses derived from this central sense, as in Sense 4 ―the location of something‖ which still belongs to the PROTO-SCENE cluster in Ho‘s semantic network of on, the data in NCCU accounts for higher percentage in literal expressions (24%) than those in BNC (13%). This suggests that learners learn on as a locational preposition and use it widely as senses under the PROTO-SCENE cluster, as in Example (4.1).
(4.1) One day Stephen goes for a walk on the beach, where he sees a young girl.
(NCCU_E006011)
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
62
In (4.1), the figure Stephen may move from one point to another in that location and beach is the place for people to act. In this way, on as a locational preposition is used to refer to the location of the figure (this is the construction of Sense 4, not Sense 1 because the support meaning is not that prominent in this construction), which is a commonly seen phrases in the data of on.
As shown in this table, some of the senses are more metaphorical, and native speakers tend to use more metaphorical expressions of on than L2 learners do in certain senses, as in Sense 13 (50 hits (5%) in BNC and 7 hits (1%) in NCCU), Sense 14 (58 hits (6%) in BNC and 15 hits (2%) in NCCU), Sense 18 (56 hits (6%) in BNC and 6 hits (1%) in NCCU), and Sense 25 (48 hits (5%) in BNC and 20 hits (2%) in NCCU). The differences in sense distribution will be helpful in comparing the semantic features in nouns constructing prepositional phrases and in describing the specific semantic profiles each preposition denotes. Moreover, the differences between native speakers‘ and L2 learners‘ command or performance of using these metaphorical constructions of on may be utilized further in finding out learners‘
problems in correctly using English prepositions.