國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
CHAPTER 4
NOMINAL GROUPS IN THE OF-CONSTRUCTION
The of -construction and its constituents, including N1 and N2 head nouns and their premodifiers, (e.g., the main cause of social differentiation) were analyzed based on the categorization schema presented in Chapter 3. The analysis of the semantic categories of the head nouns at N1 and N2 allows us to contrast their roles played in the construction. At the same time, the analysis will also reveal the inadequacy of using semantic categories to analyze the of -construction. While it will be argued that semantic relations allow a better grasp of the of -construction, an investigation into the number and type of premodifiers allows us to better capture the actual usage of the two nominal groups in the of -constructions by academic writers. A number of statistical analyses were used to identify the interaction among the factors mentioned above.
The rest of the chapter is organized in the following way. Section 4.1 presents a general distribution pattern of N1 and N2 based on their semantic categories. Section 4.2 describes the results of the correspondence analysis which was used to determine the closeness between each category. Section 4.3 and 4.4 present the distribution results of semantic relations and premodifiers, respectively. Section 4.5 presents the result of covarying collexeme analysis and Section 4.6 is on multiple of -constructions. Finally, this chapter will end with a brief summary of the findings from this study in Section 4.7.
4.1 Frequency Distribution of the Nominal Groups in the Of -construction To see to what extent the two nominal heads at both N1 and N2 would vary, their semantic categories were first analyzed. There are 14 categories in total and a total of 4,881 instances extracted from the BNC academic subcomponent were annotated as illustrated in Figure 4.1 as a bar graph. The graph compares the breakdown of 14 semantic categories of N1 and N2 head nouns arranged according to the highest N1 (‘act’) to the lowest (‘event’). Although the semantic categories of both N1 and N2 are listed in adjacency, their frequencies are independent of each other, i.e., they are not considered as a pair. The frequency of each category is presented on top of each
‧
column. While there are some minor differences between N1 and N2, there are also several striking differences between the two groups as illustrated in the graph. First, the frequency of N1 heads in ‘attribute’ is about 8.6% higher than that of N2, followed by
‘cognition’ with a smaller difference of 3.6%. In contrast, there are far more N2 heads associated with categories like ‘technical/concrete’ and ‘animate’ where the discrepancy is as wide as 11.2% and 5.3%, respectively. The rest of the ten categories namely, ‘act’,
‘group’, ‘communication’, ‘state’, ‘relation’, ‘location’, ‘process’, ‘time’, ‘quantity’, and ‘event’, do not show a variation of greater than 3%.
22.0%
Figure 4.1 Proportion of semantic categories of nominal heads in the of -construction.
To find out if the variation between the distribution pattern of the nominal groups would attain a statistical significance, a chi-square test was used to analyze the degree of divergence between N1 and N2 head nouns. A significant difference was found between these two groups (Pearson’s chi-squared = 772.64, df =13, p-value < 2.2e-16).
A further step was taken to inspect each semantic category by computing the standardized residuals (R) for each cell (see Table 4.1).
According to Haberman (1973, cited by Hinkle, Wiersma, & Jurs, 2003: 552), when the standardized residual R value for a category exceeds plus/minus 2.00, the category is a major contributor to the statistical significance. In other words, we would find categories such as ‘technical/concrete’, ‘attribute’, ‘animate’, ‘relation’, ‘process’,
‘cognition’, ‘group’, ‘quantity’, ‘communication’, and ‘state’ all contribute to the quantitative differences between N1 and N2 heads. The positive sign indicates an
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Table 4.1 The standardized residuals (R) for the semantic categories of N1 and N2 heads
Semantic categories N1 N2
Technical/concrete -12.275188 12.275188
Attribute 9.200267 -9.200267
Animate -7.144263 7.144263
Relation 6.417436 -6.417436
Process 3.528211 -3.528211
Cognition 3.327453 -3.327453
Group -3.073926 3.073926
Quantity 2.785618 -2.785618
Communication -2.311402 2.311402
State 2.005944 -2.005944
Location -1.853123 1.853123
Act 1.612923 -1.612923
Time 1.200038 -1.200038
Event 1.094541 -1.094541
excess, whereas the negative sign indicates the opposite. The results suggest that the occurrence of these semantic categories is disproportionated. For example, under the category ’technical/concrete’, the difference between N1 and N2 is the greatest (11.2%) with an R value as high as 12, indicating that there would be a higher chance of encountering a technical/concrete noun at N2 than N1. In contrast, the remaining four categories, ‘location’, ‘act’, ‘time’, and ‘event’, did not attain statistical significant difference, indicating that the chance of encountering these categories is the same for both N1 and N2. For example, the chance of encountering an ‘act’ noun is equal for both N1 and N2.
In addition to the quantitative results, the difference between N1 and N2 can also be illustrated by juxtaposing the same nominal head found at both positions from the corpus. For each category, the same nominal head (shown in bold) was chosen to provide a contrast in terms of its position (i.e., at N1 or at N2). For instance, under the category
‘technical/concrete’, the nominal head cells is illustrated in (4.1a) and (4.1b) at N1 and N2, respectively. In (4.1a), N2, the patients, appears to signal where cells come from, while in (4.1b) the word distribution refers to the frequency of occurrence of cells over
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
a space. Although these two examples both involve the same lexical head, cells, they create a different meaning with the other nominal group as they are positioned differently in the construction.
(4.1) a. The frequency of gap junctions in the gastric surface mucous cells of the patients with gastric ulcers and of healthy volunteers is shown in Table II. (H79-2049) [‘technical/concrete’ at N1]
b. To discover if an upward shift of the proliferative compartment actually takes place, we examined the distribution of labelled cells and the labelling index of the background colonic mucosa in FAP and SCRC cases by ex vivo autoradiography. (HWT-67) [‘technical/concrete’ at N2]
Examples like (4.2a) and (4.3a) demonstrate a sense of possessum at N1 (e.g., the ability and a member) and possessor at N2 (e.g., the school librarian and teacher and a firm). In contrast, when these head words are positioned at N2, different semantic relations are formed. Whereas (4.2b) describes an objectified process, i.e., the nominalization of acquiring an ability, (4.3b) describes a qualitative aspect of N2 (individuality).
(4.2) a. Figure 5.3 demonstrates the ability of the school librarian and teacher to discuss different layouts of information to meet the special needs of a specific group of pupils. (JXK-886) [‘attribute’ at N1]
b. An aptitude is demonstrated by success in learning which results in the acquisition of an ability. (CLP-1074) [‘attribute’ at N2]
(4.3) a. Under section 265 the English court has jurisdiction, for example, over a debtor who is a foreign national who has never lived or been here so long as, at a time within the last three years, he was a member of a firm which carried on business in this country. (FD8-263) [‘animate’ at N1]
b. A not uncommon solution to the feeling that the conventional family stunts the individuality of its members is for people to elect to live on their own. [‘animate’ at N2]
Similarly, example (4.4) represents a prototypical part-whole relation by involving the word part which is referred to as a focus noun (Sinclair, 1991: 87) of a following headword, i.e., focusing on the details such as parts or quantity belong to the headword at
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
not been considered to be as important as its following nominal or the head of the entire of -construction unless it is premodified as in a minor part in (4.4a). Rarely discussed in the literature is the situation in which such focus words are to be positioned at N2 as in (4.4c). Our corpus data shows that when part is used at N2 and is no longer postmodified by another of -phrase (e.g., a schematic impression of part of a computer database), it could be presented in the plural form with a sense that denotes concrete objects (e.g., body parts in the case of (4.4c)). In other words, the meaning of the word part in (4.4a) and parts in (4.4b) differs from that of parts at N2 as found in (4.4c).
(4.4) a. The vegetation of the field was analysed by ordination and correlation techniques which showed that only a minor part of the variation in species distribution could be accounted for by underlying edaphic factors… (CMA-300) [‘relation’ at N1]
b. When they are finally allowed to return a year and a half later, they discover that Xorandor and seventeen of his children are under close surveillance in different parts of the world, but that he has [programmed] them to escape and neutralize nuclear missiles.
(G1N-1501) [‘relation’ at N1]
c. In such highly evolved animals as insects the primitive segmentation, in so far as it affects the internal anatomy, has undergone profound modifications; the segmental repetition of parts is nevertheless retained to some extent in the central nervous system, the heart, tracheal system and in the body musculature. (EVW-178) [‘relation’ at N2]
Example (4.5) illustrates another instance of the head noun process. The word process at N1 in (4.5a) has been considered as an of -apposition (Keizer, 2007: 73) and signaling noun (Flowerdew, 2003; Flowerdew & Forest, 2015) where both N1 and N2 refer to an identical referent. Keizer (2007) explicates that there are two ways to look at an of -apposition. In Keizer’s example the city of San Francisco, it can be rephrased as San Francisco is a city or The city is San Francisco, depending on which nominal to start with. A similar identifying pattern can be applied to the of -construction in (4.5a) where we can rephrase it as The process is a social transformation or Social transformation is a process. The word process appears to have a labelling function (Francis, 1994). In contrast, when the word process serves as the head at N2, shown in (4.5b), a different meaning is evoked. The word process here becomes an anchor (or
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Landmark in cognitive grammar terms) that allows for extensions to be derived from itself. These two examples seem to create different meanings depending on their positions in the of -construction.
(4.5) a. Indeed; for Morris and other Marxists the nature of work necessarily changes in the process of social transformation. (A6U-1057) [‘process’
at N1]
b. Cognitive mechanisms managing intelligence are extensions of self-regulating processes in living organization. (AMG-618) [‘process’ at N2]
From the observation above, we argue that the meaning and/or function of each head nominal is highly dependent of its position, i.e., N1 or N2, in an of -construction and its relation to the other nominal head.
4.2 Correspondence Analysis of Semantic Categories of N1 and N2 Heads In the previous section, we have explored the general distribution patterns of N1 and N2 nominal heads on the basis of 14 semantic categories. However, a quantitative measurement cannot inform us much about the possible underlying relationships among them. This section presents the results of correspondence analysis (CA) which explores the possible associations among various semantic categories of N1 and N2 heads. As introduced in Chapter 3, CA is one of the multivariate methods that provides a geometric approach to categorical data. We would expect the CA analysis to identify any statistically significant associations between the N1-N2 pair. For example, the association between N1-‘animate’ and N2-‘location’ such as personal titles (e.g., the Count of Flanders, the Earl of Nottingham) or between N1-‘group’ and N2-‘location’
such as organization titles (e.g., the Bank of England), and between N1-‘cognition’
and N2-‘technical/concrete’ to express technical jargons (e.g., the concept of mankind, the pharmacology of tetracycline). Unfortunately, we did not find the expected patterns from the results. Figure 4.2 is the resulting biplot of CA with the first two dimensions contributing to an Eigenvalue of 78.20% which is calculated by adding the values given at the x-axis (51.83%) and y-axis (28.37%). The Eigenvalue is also
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
referred to as the explaining inertia, or the percentage of the structuring that explains the plot. The higher the value where the better explanatory power the plot represents.
As pointed out by Glynn (2014: 461), an explaining inertia of the first two dimensions that is more than 75% is sufficient for interpretation. As a general rule, points farther away from the central axis (point (0,0)) correspond to a higher value converted from the rows and columns of frequencies (Glynn 2014: 445). In contrast, points that are closer to the central axes correspond to a smaller value. Similarly, points that are close together represent their similarity in co-occurrence. As a consequence, on interpreting the CA biplot in Figure 4.2, some points were not considered at all. This biplot can be viewed from its four quadrants divided by the x- and y-axes. First of all, at the upper right quadrant, we can see a close association of N1-‘animate’ with N2-‘spatio-temporal’ and that of N1-‘spatio-temporal’ with N2-‘group’. The former is the most distinguishable pair as compared with the rest in the plot. Next, at the bottom-right quadrant we can find another cluster which consists of merely two groups: the association of N1-‘technical/concrete’ with N2-‘technical/concrete’. These two groups are farther away from other points, but are closer to each other than to the others, suggesting that their association would be weaker than those identified earlier.
The remaining points are all on the left two quadrants. These points are more difficult to interpret as they are nestled closely. One concern for interpreting these points is that they are very close to the center, indicating a lack of uniqueness to be distinguished from other points.
The CA provides us with a few clues, though limited, as to which categories are more closely associated with than others as summarized in Table 4.2. These three pairs represent the most prototypical of -constructions, i.e., proper names of personal titles (e.g., the new Republican President of the United States) and locative titles (e.g., the Bank of England). Technical concrete nouns like a piece of paper would also be expected as they represent the topics to be discussed or reported in academic writing.
However, it was surprising that none of the ‘act’ nouns at either N1 or N2 were found to have an association with another group, even though the ‘act’ nouns were high in
‧
Figure 4.2 Correspondence analysis of semantic categories of nominal heads.
proportion. On the whole, there were only a handful of distinctive N1-N2 associations identified by using CA. As far as categorization of the of -construction is concerned, semantic categories of N1 and N2 seem to be insufficient to capture the various types of of -constructions. This could be due to the large number of possible combinations of N1 and N2 semantic categories (14 x 14 = 196) found in the of -construction. After the deletion of cells with a low frequency to ensure the accuracy of the analysis, a total of 100 combinations were subject to CA. It is possible that our annotated data might not have been sufficient in amount to attain additional significant results. In the next section, we analyze the of -construction based on their semantic relations.
4.3 Semantic Relations of the Of -construction
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Table 4.2 A summary of CA results with examples Semantic categories
N1 N2 Example
animate spatio-temporal the new Republican President of the United States (A66-805)
the rival theorists of the pre-science period (FBE-1531)
Lord Slynn of Hadley (FDK-78) spatio-temporal group a chosen area of the country
(HJ0-11334)
back of the queue (FPJ-969)
the post-war demise of working-class communities (G1H-648)
the Concordat period of Roman catholic church (A07-696)
technical/concrete technical/concrete three images of the gall bladder (HU2-4463)
the circuits of a computer (EUS-250) a piece of paper (FD2-95)
the products of the furnace (EEM-809)
We believe that the most unique feature of the of -construction is perhaps its inherent nature in expressing the semantic relation between N1 and N2. This section reports the distribution pattern of various semantic relations annotated according to the procedures described in Section 3.4.3. A total of four major categories (‘part-whole’,
‘quality’, ‘action’, and ‘derivative’) and 17 sub-categories were annotated. The category ‘part-whole’ consists of five sub-categories, namely, ‘belonging’, ‘partitive’,
‘spatio-temporal’, ‘type’, and ‘relational’. The second category ‘quality’ comprises of six sub-categories, namely, ’attributive’, ‘topic’, ‘equative’, ‘measurement’,
‘collective’, and ‘containment’. The third ‘action’ category has only three sub-categories: ‘subjective’, ‘objective’, and ’others’. The final ‘derivative’ category also has three sub-categories: ‘product-source’, ‘resultative’, and ‘causal’. Table 4.3 presents the occurrence rate of each type of semantic relations annotated from the 4,881 instances of concordance with examples. Overall, the most common category is
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
the ‘quality’ relation (44.61%), followed by ‘part-whole’ (24.96%), ‘action’ (22.50%), and then ‘derivative’ (8.00%). When we examine closely at the sub-categories from the table, by far the highest frequency for the sub-categories is ‘objective’, followed by
‘attributive’, and ‘topic’. Under the main category of ‘action’, the ‘objective’ category mainly consists of deverbal nouns (e.g., creation, assessment) and their participant under the thematic role of a patient (e.g., industry, pupils).
In this section, our focus is on ‘bare’ nominals only, i.e., nominal groups without any premodification in the of -construction, leaving premodification to be discussed in Section 4.4. A total of 2,059 instances or 42.2% were identified as exemplified in (4.6).
(4.6) a. The ultimate goal was the creation of an industry, originally funded by foreign investment... (B1E-273)
b. The results will further the development of measures of school effectiveness and of performance indicators, and will assist in evaluating the impact of recent changes in the assessment of pupils.
(HJ1-16104)
c. Socialization theories offer the possibility of shaping and changing behaviour, including your own. (CMR-1094)
d. Provision should be made for the chairing of each meeting and for the taking of votes. (J6P-634)
Out of the 2,059 instances mentioned above, there are about 11.9% (243 instances) of nominal heads found as verbal nouns in the (-ing) form. As Examples (4.6c) and (4.6d) show, these action nominals can be found in both N1 (5.3%, 110 instances) and N2 (6.6%, 136 instances) with an approximately even proportion from the data. The nominal heads at N1 include words such as means, task, way, cost, difficulty, purpose, mode, matters, methods, point, power, possibility, principle, and problem, all of which belong to the group of shell nouns (Hunston & Francis, 2000; Schmid, 2000; Aktas &
Cortes, 2008) or signaling nouns (SN) (Flowerdew, 2003; Flowerdew & Forest, 2015).
This pattern of [N1 of V-ing] is in congruence with Flowerdew and Forest’s (2015) work on SNs under the syntactic pattern [SN + of + ing-clause]. In their work, the top five SNs accounting for over 50 per cent of occurrences are way, idea, chance, possibility, case, and process. Flowerdew and Forest’s research is mainly concerned with genral and
‧
Table 4.3 Frequency distribution of semantic relations of of -constructions Semantic
relation
Subcategory Example Frequency
part-whole
belonging the rights and duties of the individuals (ABP-653)
367 7.52%
1218 (24.96%)
partitive the centre of the Earth (H8K-181)
341 6.99%
spatio-temporal the census of 1981 (HHY-2899)
279 5.72%
type a mode of production (F9G-1392)
202 4.14%
relational an ally of the US (EFA-798) 29 0.59%
quality attributive pattern of behavior (EDH-514)
659 13.50%
2178 (44.61%)
topic the expression of racism (FA9-1236)
496 10.16%
equative an act of faith (CMN-1068) 415 8.50%
measurement one piece of information (GVA-814)
427 8.75%
collective a list of values (A07-159) 110 2.25%
containment a form of thought (G0R-565) 71 1.45%
action objective the administration of justice (EDW-188)
970 19.87%
1096 (22.50%)
subjective the approval of the church (EDF-725)
124 2.54%
others the recipient of the relevant benefit (J7A-1222)
2 0.04%
derivative product-source the decision of the Hight Court (FBJ-666)
138 2.83%
261 (8.00%)
resultative analysis of variance (HWU-473)
78 1.60%
causal the causes of epidemic (HXF-2145)
173 3.54%
Total 4,881 100%
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
disciplinary variation. They found that this pattern is particularly prevalent in lectures (44.0%) as compared to journals (29.3%) and textbooks (26.7%) and they point to a wide disciplinary gap between natural sciences (24.8%) and social sciences (75.2%). In our corpus data, we found an interesting but complex embedment within this type of construction as shown in (4.7).
(4.7) It may take the form of denying the normally expected clues of context and coherence... (EWA-243)
This construction has two of -constructions but within two different embedment. The second of -construction, the normally expected clues of context and coherence, is embedded in the verbal gerund denying, which forms the first of -construction with the form. This type of complex constructions appears to be not uncommon in our corpus, exemplifying a multi-layer embedment within a noun phrase.
The results in Section 4.3 provide a general descriptive statistics for the frequency distribution of the four major types of semantic relations, namely, ‘part-whole’
(24.96%), ‘action’ (22.45%), ‘quality’ (44.61%), and ‘derivative’ (7.97%). It was also showed that the ‘quality’ relation remains to be the most popular type in the of -construction. Yet, if we compare across the sub-categories, ‘objective’ relation is
(24.96%), ‘action’ (22.45%), ‘quality’ (44.61%), and ‘derivative’ (7.97%). It was also showed that the ‘quality’ relation remains to be the most popular type in the of -construction. Yet, if we compare across the sub-categories, ‘objective’ relation is