A diachronic perspective on near-synonymy: The concept of sweet-smelling in American English

Daniela Pettersson-Traba* A diachronic perspective on near-synonymy: The concept of SWEET - SMELLING in American English https://doi.org/10.1515/cllt-2018-0025 Abstract: This paper presents a diachronic analysis of the attributive uses of four synonymous adjectives which designate the concept of SWEET - SMELLING (fragrant, perfumed, scented,andsweet-smelling) in the latter part of Late Modern and Presentday American English. By drawing on data from the Corpus of Historical American English (COHA) and applying a Hierarchical Configural Frequency Analysis (HCFA), it delineates the internal semantic structure of this set of synonyms, paying special attention to their noun collocates. The results show that the concept of SWEET - SMELLING experiences major changes over the time span examined (1850–2009), from being used mostly to qualify entities which can exhibit a natural pleasant smell (e.g. flowers and trees) to modifying objects which are artificially sweet-smelling (e.g. oils and shampoos). Moreover, fragrant and perfumed, which initially were the most frequent adjectives, are gradually replaced by scented, thus reflecting a change in the relation between the synonyms over time. The study constitutes the first diachronic approximation to synonymy from the perspective of cognitive semantics and provides equally effective results as previous synchronic research in the field. Keywords: synonymy, semantic change, collocational behavior, Hierarchical Configural Frequency Analysis 1 Introduction Although absolute synonyms are virtually nonexistent, languages abound in near-synonyms, linguistic constructions with similar, though not identical meanings. 1 Near-synonyms are fundamental to the organization of the lexicon *Corresponding author: Daniela Pettersson-Traba, Department of English and German, University of Santiago de Compostela, Santiago de Compostela, Spain, E-mail: [email protected] http://orcid.org/0000-0001-9302-3372 1Throughout this paper I employ the terms synonym/synonymy and near-synonym/nearsynonymy interchangeably since many linguists concur on absolute synonymy being very infrequent in languages, if it exists at all. Corpus Linguistics and Ling. Theory 2021; 17(2): 319–349 Published online October 09, 2018 of any language since they exert a great influence on the structure of lexical knowledge. For instance, synonyms lie at the basis of the creation and demarcation of meanings due to their contrastive nature. As claimed by Saussure (1983: 114), [i]n a given language, all the words which express neighbouring ideas help define one another’s meaning. Each of a set of synonyms […] has its particular value only because they stand in contrast with one another. […] So, the value of any given word is determined by what other words there are in that particular area of the vocabulary. To be considered synonymous, words or larger linguistic units must refer to the same concept in the extralinguistic reality, that is, they must share core denotational semantic traits. However, synonyms are permitted to contrast in peripheral aspects of their denotational meaning, or in other dimensions such as style, connotation, and collocation. Thus, synonyms typically convey different nuances and implications, making it difficult to choose the most suitable word from a group of near-synonyms in specific contexts of use. Despite its complexity and importance, research on synonymy is scarce, especially if compared to other semantic phenomena such as polysemy and metaphor, which have been the subject of an ever-increasing body of literature (cf. Edmonds and Hirst 2002; Taylor 2003; Liu 2010). The semantic structure of specific groups of synonyms has received particularly little attention, and several researchers have recently stressed the need for more studies in this field. For instance, Divjak and Gries (2006) point out that “the issue of the internal structure of a set of synonyms has hitherto remained largely undiscussed in the literature”, thus pointing to a clear existing gap in this area of research. Liu also draws attention to this fact in several of his studies on synonymy (Liu 2010, 2013; Liu and Espino 2012), claiming that “[a]dditional systematic research is needed to share more light on this intriguing linguistic phenomenon and, more importantly, to gain a better understanding of how various specific sets of synonyms work in terms of meaning and usage patterns”(Liu 2010: 57). This has led to the emergence in the last couple of decades of studies following current distributional methods such as Behavioral Profiles (cf. Gries and Divjak 2009; Gries 2010) and word-space models (cf. Heylen et al. 2008; Levshina and Heylen 2014), which focus precisely on the semantic structure of specific synonyms. Even though these studies have all successfully uncovered subtle distinctions in meaning between synonyms by analyzing, for instance, their semantic, syntactic, and/or stylistic behavior, they have only dealt with the semantic structure of synonyms from a synchronic viewpoint, while their diachronic evolution generally has been neglected. The present paper seeks to partially 320 Daniela Pettersson-Traba fill this gap by focusing on the distributional patterns of a set of adjectival nearsynonyms in the history of American English. By adopting a distributional orientation, this study examines the attributive uses of four adjectival near-synonyms, namely, fragrant, perfumed, scented, and sweet-smelling, in the latter part of Late Modern (LModE) and Present-day American English (PDE) (1850–2009). 2 This is done by drawing on data from the Corpus of Historical American English (COHA) (Davies 2010-) and by carrying out a Hierarchical Configural Frequency Analysis (HCFA). 3 The focus is on the collocational behavior of the synonyms, and more specifically on the semantic categories of their head nouns. There are multiple reasons for deciding on this group of synonyms. First, to the best of my knowledge, no previous study on synonymy has focused on lexical items within the domain of smell. In fact, a common claim in the literature is that people are bad at naming smells (cf. Henning 1916; Lorig 1999; or Yeshurun and Sobel 2010; among others), which is probably related to the fact that the odor vocabulary has not been as widely researched as other domains such as that of color (cf. for instance Berlin and Kay 1969; or Biggam 2012) or cookery (e.g. Lehrer 1969, 1974), which have been argued to be much more extensive and precise in English. However, the Historical Thesaurus of the Oxford English Dictionary (HTOED)lists 894 words under the semantic category SMELL AND ODOR , which have been and/or are used in English, a fact which appears to contradict previous claims. Of these words a substantial amount are adjectives similar to those selected in the present study with some minor variations in definition or use, making this a particular interesting semantic field for the study of near-synonymy. Second, the four synonyms appear to be interchangeable as they are defined in practically the same way, as “having a pleasant smell”in existing reference material (Cambridge Dictionary, Collins Dictionary,English Oxford Dictionaries, Longman Dictionary of Contemporary 2Over the course of this paper, sweet-smelling refers to the adjective, whereas SWEET - SMELLING is used to refer to the whole concept, as denoted by the four synonymous adjectives (i.e. fragrant, perfumed, scented, and sweet-smelling). Additionally, SMALL CAPITALS are used for semantic categories and conceptual metaphors, while variables and meanings are placed within single quotation marks (“”). Dictionary citations are structured as follows: (OED s.v. breath noun 3a). The first element indicates the name of the dictionary in abbreviated form, followed by s.v, which stands for sub voce “under the word”, head word, part of speech, and finally sense number. 3The HCFA is a multivariate technique used for categorical variables which analyzes the simultaneous effect of various factors on a particular phenomenon. This test aims at identifying interactions between the variable and variable-levels included in the analysis in order to detect types or groups with similar behaviors or characteristics (Hilpert 2013: 56–57). For more details, see Section 3.2. A Diachronic Perspective on Near-Synonymy 321 English,Merriam-Webster,Oxford English Dictionary (OED), The American Heritage Dictionary of the English Language). 4 Additionally, some dictionaries employ circular definitions, i.e. definitions in which the synonyms are defined in terms of each other. Third, the examples of usage provided do not clarify which sort of nouns they commonly co-occur with. Several nouns such as flower, soap,andsmell appear modified by more than one of the adjectives without any apparent change in meaning. Only some of the dictionaries consulted offer information about their particular nuances of meaning: three of the eight dictionaries consulted provide additional senses for perfumed (English Oxford Dictionaries,Collins Dictionary,OED) and scented (Collins Dictionary,OED); none for fragrant or sweet-smelling. According to these dictionaries, there seems to be a difference in nuance in the case of the former two adjectives depending on whether the source of the smell is natural or artificial. This difference is reflected in the kind of nouns that the adjectives modify. In the case of natural sources of smell, perfumed and scented modify nouns, as for instance, flower, coffee,andsea, which can release a natural sweet pleasant smell, whereas in the case of artificial sources, they collocate with nouns such as dress, cream,andsoap, which can acquire an agreeable smell only by being impregnated by a sweet-smelling substance. Yet, the information is limited, and it is impossible to identify in which contexts the synonyms can be used interchangeably. This lack of adequate and comprehensive coverage of the semantic properties of the four synonyms calls for a more detailed analysis. The remainder of this article is organized as follows: Section 2 provides an overview of the theoretical foundations and developments of the distributional method to lexical semantics, and more specifically to the analysis of synonyms. Section 3 focuses on the data retrieval, methodology, and statistical analysis proper, whereas Section 4 puts forward the results obtained as well as a discussion of their implications. Finally, Section 5 offers some concluding remarks and suggestions for future research. 4The selection of the four near-synonymous adjectives is based on careful examination of the definitions and examples of usage provided by these reference materials. While it is true that some other adjectives share the same definition as the selected ones (e.g. fragranced) or are listed as their synonyms (e.g. aromatic) and could thus have been chosen as well, these lexical items have been excluded for different reasons. The adjective fragranced is classified as rare in several of the reference materials and it is not attested in COHA. The definition of aromatic differs slightly from the adjectives under scrutiny and does not necessarily imply the trait sweetness. Additionally, many of the other adjectives listed in the HTOED are much less frequent than the four items examined here. Nevertheless, a selection of a wider range of near-synonyms, including adjectives such as sweet-scented, sweet (OED s.v. n.6), balmy (OED s.v. n.3), or redolent, might prove useful or even necessary in future studies. 322 Daniela Pettersson-Traba 2 Literature review In the last decades, there has been a particular interest in the semantic structure of synonyms, which has primarily been due to scholars who follow a contextual approach to semantics. However, the distributional method for lexical semantics has a long history, going back to the neo-structuralist school. According to this method, the semantic and functional features of lexical items closely correlate with their distributional patterns. More specifically, this tradition rests on three major tenets: (1) its radical usage-based orientation, (2) the paramount importance of collocation, i.e. the “lexical relation between two or more words which have a tendency to co-occur within a few words of each other in running text”(Stubbs 2001: 24), and (3) its strong link with statistical analysis. Scholars such as John Firth (1957a, 1957b) and John Sinclair et al. (1966, 1987, 2004) were among the first to suggest that distributional patterns and collocations are essential for the analysis of lexical semantics. However, it was not until the latter half of the twentieth century that major advancements in corpus research took place which allowed these scholars’view of thinking to progress considerably. Association measures such as the Pointwise Mutual Information index (PMI) (Church and Hanks 1990) became groundbreaking as they permitted to quantify the probability of association between collocates, and have since been widely employed. Additionally, further statistical techniques such as regression analysis and clustering methods, e.g. hierarchical agglomerative cluster analysis (cf. Gries 2009: 306–319) and conditional inference trees (cf. Levshina 2015: 291–300) have led to further development in the field. Concerning synonymy, in the 1990s and early 2000s several studies paid close attention to the collocational and colligational patterns of near-synonyms (i.a. Kennedy 1991; Church et al. 1994; Biber et al. 1998: Section 2.6; Partington 1998: Section 2 and 3.6; Kjellmer 1995; Taylor 2003), employing association measures like the PMI or the t-score (Gries 2001, 2003). 5 Despite their valuable findings, which have contributed to a better understanding of the behavior of synonyms, there is still room for polishing this method. In fact, Gries (2010) highlights three areas which could be fine-tuned. First, most of the aforementioned studies focus on pairs of synonyms (cf. Kjellmer 1995; on almost and nearly; Taylor 2003; on high and tall), whereas the majority of near-synonyms come in larger sets. Second, the range of distributional patterns considered is 5Colligations are defined as “the collocation of a node word with a particular grammatical class of words”(McEnery et al. 2006: 82). A Diachronic Perspective on Near-Synonymy 323 rather limited, being restricted either to collocational or colligational behavior. Last, some studies lack explanations of the implications of the results obtained and do not relate them to any linguistic or semantic theory. In view of this backdrop, scholars within the context of cognitive semantics, have recently employed innovative methods to the study of synonymy (see Section 1) with exceptionally positive results, thus proving these methods to be particularly effective in delineating the fine-grained aspects of meaning and in unfolding the differences existing between synonyms. Most research in this domain have concentrated on sets of near-synonymous verbs (cf. Divjak 2006; Divjak and Gries 2006; Speelman and Geeraerts 2009; Levshina 2011). In a Behavioral Profile study, Divjak and Gries (2006) examined the distributional patterns of nine Russian verbs of trying to analyze their syntactic and semantic patterning. The syntactic variables include, for instance, verb related characteristics such as aspect and mode as well as clause and subject related characteristics such as sentence type and clause structure. Semantic variables comprise aspects concerning the nominative subject paradigm, that is, oppositions such as animate vs. non-animate or concrete vs. abstract. Moreover, they analyze the adjacent verb collocates of the synonyms by grouping them into semantic categories such as PHYSICAL ACTION , PHYSICAL PERCEPTION , SPEECH , and INTELLECTUAL ACTIVITY . This was done with the intention of unfolding their primary and secondary meanings and differentiating them from their respective synonyms. Divjak and Gries’study is just one of the studies which have proven the usefulness of the Behavioral Profile approach in facilitating the identification of the semantic structure of verbs and can be said to have paved the way for further and more detailed research along the same lines on other parts-ofspeech: adjectives (cf. Gries and Otani 2010; Liu 2010), nouns (cf. Liu 2013), and adverbs (cf. Liu and Espino 2012). Concerning research on near-synonymous adjectives, we find Gries and Otani (2010) on big, great, and large and their respective antonyms (i.e. little, small, and tiny), and Liu (2010) on chief, main, major, primary, and principal. Liu delineated the semantic structure of the five synonyms by drawing on data from the Corpus of Contemporary American English (Davies 2008–) and carrying out a HCFA. He discovered significant differences between the attributive adjectives by measuring their collocational (i.e. the semantic categories of nouns modified by the synonyms), colligational, and stylistic preferences (i.e. text type). The reasons for paying close attention to the head nouns are well justified, since the existing literature has demonstrated that looking at the nouns with which adjectives collocate is one of the best ways to reveal the nature of the semantics of adjectives. For instance, Geeraerts’(1986) study on the Dutch adjective vers “fresh”demonstrated that the fine-grained aspects of meaning of polysemous 324 Daniela Pettersson-Traba adjectives can be discovered by examining their head nouns. Geeraerts provides evidence for vers having different meanings depending on the nouns it accompanies. For instance, when it modifies wound, vers means “recent”, but when it modifies the noun air, it takes on a slightly different meaning, namely “pure, untainted”,or“optimal”. Gries’(2001, 2003) studies on English –ic and –ical adjective pairs, e.g. economic and economical, has also contributed interesting findings by focusing on the head nouns of the paired adjectives. Liu (2010), who classified the head nouns into six semantic categories, to wit CONCRETE , ABSTRACT , DUAL , INSTITUTION , POSITION TITLE , and NON - POSITION TITLE , learnt that whereas dual and abstract nouns were often modified by the five adjectives, the other semantic categories tended to be dominated by one or two of them: for instance, while concrete nouns preferred main, position title nouns were dominated by chief. As demonstrated above, multiple studies have been carried out with effective results, particularly within the context of cognitive semantics. In fact, this theory converges with the original neo-structuralist distributional method in many ways, not the least due to their reliance on actual usage-based data and their quantitative orientation. Given the rapid increase of studies in cognitive semantics over the last two decades, its importance and utility cannot be underestimated (cf. Geeraerts 2010: 264–265). This brief review has demonstrated the effectiveness of the distributional corpus-based approach in uncovering fine-grained semantic and usage distinctions between sets of synonyms as well as the motivations for continuing with research along the same lines. This is precisely what the present study aims at, and the findings obtained offer a significant contribution to the field. It does so by considering the semantic category of head nouns, and by presenting a more detailed semantic classification than most previous studies. More importantly, the approach adopted differs from earlier research in that the diachronic development of the synonyms is also accounted for, somethingwhichtomyknowledgehasnotbeendonebefore.Themain research questions addressed are: (1) Do the near-synonyms appear with diverse semantic categories of nouns? If so, are the differences significant? (2) Are there any major changes taking place over the time span analyzed which have an influence on the use of the concept SWEET - SMELLING ? If so, where are these changes taking place and what are the implications? (3) Could the innovative methods put forward in the context of cognitive semantics be applied to the diachronic study of near-synonyms with equally effective results as in previous synchronic research? A Diachronic Perspective on Near-Synonymy 325 3 Data and method To provide a detailed description of the synonyms selected, fragrant, perfumed, scented, and sweet-smelling, the first and most elementary information required is their overall frequency. However, this study focuses only on the attributive uses, and more specifically on those uses in which there is no intervening material between the synonymous adjectives and the head nouns. Therefore, it was necessary to discard the uses where the synonyms function as verbs or as predicative complements, as well as the attributive uses in which the head nouns did not directly follow the adjectives. 6 Consequently, this study is based on 1,666 instances of the near-synonyms. This section describes the process of the data retrieval, the semantic classification of nouns (Section 3.1), and the statistical analysis (Section 3.2). 3.1 Corpus: data retrieval and semantic classification The data used was retrieved from COHA, which, with approximately 400 million words, suits perfectly the purposes of this analysis, as it allows us to examine low frequency items such as the synonyms under study with more reliable results than smaller corpora. However, even though COHA covers the time span 1810–2009, this study only focuses on the decades from 1850 onwards. The reason for this is that COHA is not a balanced corpus in terms of number of words, as it contains considerably more data for the later decades than for the earlier ones. In fact, until the 1850s there are less than 16,000,000 words per decade, whereas the last two decades, 1990s and 2000s, include nearly double the amount of running text. To make the decades comparable, this analysis is based on approximately 16,493,826 words from each decade (i.e. the same as the total number of the first analyzed decade, the 1850s), resulting in a total number of 263,901,889. Finally, the decades were grouped into four 40-year periods, that is, 1850–1889 (Period 1: P1), 1890–1929 (Period 2: P2), 1930–1969 (Period 3: P3), 1970–2009 (Period 4: P4), of around 65,975,466 words each, as shown in Table 1. 6Examples as the following were thus not considered in the analysis: (i) perfumed as verb: He held the shirt up. It perfumed the air rapidly. (COHA, 1943, FIC) (ii) predicative use: They were sweet-smelling, sweet as flowers out of season. (COHA, 2001, FIC) intervening material: I had picked my mother some fragrant pale pink Renaissance roses from her garden. (COHA, 2009, FIC) 326 Daniela Pettersson-Traba To acquire information concerning the nuances of meaning of the synonyms, it was necessary to identify the specific usage patterns of each adjective. To do so, I followed the methodology employed in Liu (2010). As the focus lies on the head nouns, the most common noun collocates of each adjective were retrieved by using the PMI score, which compares the likelihood of two words co-occurring with the likelihood of the individual words (Church and Hanks 1990: 23). 7 This procedure facilitated the creation of a list of approximately ten top adjacent nouns, that is, R1 collocates, for each of the synonyms in the four different periods established. 8 Given that I queried the PMI in each of the periods established, I ended up with a total of 16 lists (i.e. four for each synonym, one for each period). However, as there was a considerable degree of overlap between the lists, both between the adjectives and the different periods, the final list of top nouns contained 59 different noun types in total. These 59 nouns were subsequently grouped into nine fine-grained semantic categories based on the semantic classification offered by the HTOED, which classifies each of the senses of a given word into categories: FOOD & DRINK , PLANTS & FLOWERS , EARTH , BODY , MATTER , SENSATION , AESTHETICS , CLEANING ,and TEXTILE & CLOTHING . 9 Then, a query was conducted in COHA to retrieve all adjacent Table 1: Number of words per decade and period. PWords P Words P Words P Words  ,,  ,,  ,,  ,,  ,,  ,,  ,,  ,,  ,,  ,,  ,,  ,,  ,,  ,,  ,,  ,, Total ,, Total ,, Total ,, Total ,, Total words in subcorpus: ,, 7The PMI score was used since there is a specific query in COHA which calculates the PMI of the collocates of the target word. The PMI was used is it favors words with a generally low frequency in the corpus, but which tend to collocate with the adjectives. Additionally, The PMI was used instead of the more habitual t-score, since the PMI is more suitable when dealing with content words as is the case in the present analysis, while the t-score performs better in the case of function words (cf. Church et al. 1994 for further details on these association measures). 8In view of the low frequency of some of the adjectives, especially sweet-smelling, it was not always possible to obtain a list of 10 top nouns for each period. For this reason, I also made used of the frequency measure. This measure provides the nouns that the adjectives most frequently modified irrespective of their overall frequency in the corpus. As Liu (2010: 63) points out, “the frequency measure and the MI score can complement each other in better identifying the nouns that are typically modified by the adjectives.” 9As each of the synonyms modify a huge number of individual types of noun, it would be extremely time consuming to categorize all senses of the individual head nouns. Consequently, A Diachronic Perspective on Near-Synonymy 327 Figure 2: Distribution of the synonyms across the semantic categories. 334 Daniela Pettersson-Traba However, the remaining semantic categories are dominated by one or two of the adjectives and the results do reveal significant differences between the synonyms. In FOOD & DRINK (e.g. chicken and wine), fragrant is the only type and there are just a few examples of the other three adjectives with these nouns. What is more, both scented and perfumed are antitypes with nouns denoting food and drink. Likewise, fragrant is the only type with PLANTS & FLOWERS (e.g. geranium and leaf)whereasscented and perfumed are both antitypes. The only difference is that sweet-smelling is a considerably frequent modifier of this sort of nouns, with nearly half of its examples (49%) occurring with nouns designating plants or flowers. Contrarily, in the CLEANING category (e.g. disinfectant and soap), scented and perfumed are the only types while fragrant turns out to be an antitype with only five examples (that is, approximately 0.54% of its use). Similarly, sweetsmelling is rather infrequent with this kind of nouns as it only occurs collocating with them in four of the instances (3.7%) in the data analyzed. In the remaining categories, namely BODY (e.g. skin and hair). TEXTILE & CLOTHING (e.g. dress and silk), and AESTHETICS (e.g. cream and lotion), perfumed prevails, being the only type, while fragrant is considered an antitype in the latter two. The findings obtained show that the adjectives do, in general, have different collocational preferences as they tend to co-occur with different categories of nouns. These differences in collocational affinity demonstrate that synonyms are normally not interchangeable, but rather tend to collocate with diverse types or sets of words despite sharing the same sense (Cruse 1986: 278–281). For instance, Murphy (2003: 256) explains that although some of the existing differences between near-synonyms, for instance those relating to collocations or frequency, do not influence the communicative intent, they do provide evidence that Table 5: Distribution of the adjectives across the nine semantic categories. Adj. Cat. Fragrant Scented Perfumed Sweet-smelling FOOD & DRINK  T AA PLANTS & FLOWERS  T A A BODY    T EARTH     MATTER     TEXTILE & CLOTHING  A   SENSATION   A  AESTHETICS  A  T CLEANING A T T Total:     A Diachronic Perspective on Near-Synonymy 335 synonyms are rarely substitutable. Moreover, she adds that words typically collocate with a limited amount of other words and, despite the fact that replacing a word by any of its synonyms does not always lead to ungrammaticality, it normally results in a non-natural combination due to collocational affinities. One important picture emerges from the results obtained so far, namely a clear division of semantic labor between the synonyms regarding the two senses previously referred to (cf. Section 1): (1) “Naturally possessing or emitting a pleasant sweet smell”and (2) “permeated with a substance which artificially adds a pleasant sweet smell”. As seen in Table 2, by closely analyzing the nouns in each semantic category, it becomes clear that some of them solely comprise nouns which denote entities with a natural pleasant smell, i.e. PLANTS & FLOWERS , FOOD & DRINK ,and EARTH , whereas other include nouns which exclusively refer to man-made or artificial objects, i.e. TEXTILE & CLOTHING , AESTHETICS ,and CLEANING .Whereasfragrant is a type with nouns which denote entities with a natural pleasant smell, perfumed and scented are types with nouns referring to man-made, artificial objects, i.e. those in CLEANING (in the case of both adjectives), AESTHETICS ,and TEXTILE & CLOTHING (only in the case of perfumed). This division of semantic labor hence points to the fact that perfumed and scented might be the only adjectives which are truly polysemous, being used also in sense 2, as pointed out by some of the dictionaries consulted (cf. Section 1). The rest of the semantic categories, i.e. MATTER , BODY , and SENSATION , encompass nouns which can be used with the adjectives either in sense 1 or 2 depending on the context of use. There are certainly instances of the adjectives with these nouns in which it is clear whether they are used in sense 1 or 2, as illustrated in (6) to (8). (6) Shadow lAke’s citrus-based laundry detergent, Citra-Suds, makes short work of both those problems ($7.99). The light fruity smell lends laundry a fragrant aroma without being perfumey, and its cleansing power is dramatically boosted with an oxygenating agent called Oxi –Clean. (COHA, 2001, MAG) (7) His leaf-land is fair and so marvellous fair, His palm-land is filled with a perfumed air Of magnolia blooms to its dome of blue. His rose-land has arbors of moss-swept oak, –Gray, Druid old oaks; and the moss that sways And swings in the wind is the battle-smoke Of duellists, dead in her storied days. (COHA, 1887, FIC) (8) Things she told me, sitting on the edge of my bed at night: “I was born with a caul. That means I have a sixth sense.”Or, touching her perfumed wrist to my cheek: “This is called Shalimar.”(COHA, 2006, FIC) 336 Daniela Pettersson-Traba In these examples, the surrounding context, either intra–or extralinguistic, indicate in which sense the adjectives are used. In (6), the words in the surrounding context window of fragrant (e.g. laundry) indicate that this instance of the adjective with the sensation noun aroma is used in sense 2. In (7), the prepositional phrase of magnolia blooms and other collocates, such as leaf-land and arbors, reveal that air denotes the atmospheric air of outdoors (as opposed to the air of confined spaces) and is thus naturally pleasant smelling. Likewise, in (8), perfumed is used in sense 2, as made evident by the extralinguistic situation, since one of the characters refers to a well-known perfume, namely Shalimar, and general knowledge: people typically put perfume on the wrists. However, some occurrences of these categories of nouns are more ambiguous and it is difficult –sometimes even impossible –to know for sure whether the modifying adjectives are used in sense 1 or 2. Consider the following two examples, in which scented and fragrant modify a noun belonging to MATTER and BODY , respectively. (9) I climbed back inside and took out the magazine and studied Miss July by flashlight and thought about how her skin would feel where her breasts curved back into her body toward her underarms, and how her nipples would feel, and the hair between her legs, the lips between her legs. When I had finished my business, I lay on top of the sheets and let the warm scented air wash over me. (COHA, 1995, FIC) (10) Off came her earrings, then her clothes, without embarrassment. Breasts or none, her woman’s wealth satisfied him. When he touched her nipples the effect was electric. She could be a little wild in bed. “Your fingers are fires.”He enjoyed her long body, the cool flesh, sex smell, fragrant hair.(COHA,1961,FIC) In (9), it is clear that air refers to the air of a confined space, but it is not clear whether scented denotes a natural odor as emitted by, for instance, the body of the person in the room, or by an artificial one as emitted from, for example, detergent that the sheets mentioned have been impregnated with, or from perfumes worn by him/her. Similarly, in (10), the man might find the hair of the woman naturally fragrant, or its sweetness might be due to the shampoo or another cleansing agent. In both examples, it might even be the case that the pleasant odor results from a mixture of both naturally and artificially pleasant-smelling substances. Occurrences of the adjectives such as these cannot be categorized clearly into either sense 1 or 2 but are open to interpretation. Thus, in these three categories, there are examples in which the adjectives cannot be said to be used clearly in either sense 1 or 2 as in the previous cases (i.e. FOOD & DRINK , PLANTS & FLOWERS , EARTH , CLEANING , AESTHETICS ,and TEXTILE & CLOTHING ). A Diachronic Perspective on Near-Synonymy 337 The results presented in this section reveal important insights about the phraseological tendencies of the synonymous adjectives, as well as the similarities and differences between them. Notwithstanding, these results do not account for their distribution across the nine categories of nouns at specific time periods during the 160 years considered, for potential changes in collocational behavior, for those specific to each adjective, or for those concerning the concept of SWEET - SMELLING . This is the concern at issue in the next section, in which the collocational preferences of the near-synonyms are analyzed in four separate 40-year periods: 1850–1889, 1890–1929, 1930–1969, and 1970–2009. 4.3 Semantic development: Changes in collocational behavior To examine the collocational patterning of the synonymous adjectives in the four periods distinguished, it was necessary to cross-cut the three variables “Synonym”,“Semantic category of head noun”, and “Period”in the HCFA. Due to the relatively low frequency of the examined adjectives, it was difficult to obtain statistically significant results. However, it has been possible to determine which cells in the contingency table contribute to the significance of the results of the general collocational preferences. For instance, both perfumed and scented showed up as antitypes with nouns belonging to PLANTS & FLOWERS in the general analysis (see Section 4.2), but the results obtained when including “period”reveal that while perfumed is an antitype in all four periods with these nouns, only P3 (1930–1969) plays a part in scented being considered an antitype (cf. Table 6). Additionally, this analysis yields findings which ascertain that some major changes in collocational behavior are going on. First, PLANTS & FLOWERS nouns become substantially less frequent over time with fragrant, perfumed, Table 6: P LANTS &F LOWERS throughout 1850–2009. Adj. Period Fragrant Scented Perfumed Sweet-smelling Total P:–  T A   T P:–   A   P:–   AA T A P:–  AA   A Total:      338 Daniela Pettersson-Traba and scented. In the case of fragrant, this category goes from being modified more frequently than expected in the first two periods (a type in P1), to being modified less frequently in the later periods, and finally becoming an antitype in P4. With both perfumed and scented, these nouns always appear as collocates less frequently than expected, but a decline can still be appreciated. With perfumed, PLANTS & FLOWERS is an antitype throughout 1850–2009, but its percentage of use with the adjective decreases over time, going from 16.7% in P1 to 0.13% in P4. With scented, the category becomes an antitype in P3, but there is an overall downward tendency, going from making up 46.8% of the uses in P1 to 19.6% in P4 (cf. Figure 3). The only exception is sweet-smelling, which modifies PLANTS & FLOWERS significantly more frequently than expected in P3, thus being a type. The output of the HCFA also indicates a change in the categories CLEANING with both perfumed and scented, and AESTHETICS with perfumed (See Tables 7 and 8). The adjective perfumed always modifies nouns belonging to CLEANING more frequently than expected, but gains more ground in P3, in which it becomes a type. The percentages of use also indicate a gradual increase as this category goes from totaling 7% of the use of perfumed in P1 to 20.4% in P4 (cf. Figure 4). With scented, CLEANING nouns occur less frequently than expected in P1, but more frequently in the remaining three periods, becoming a type in P3 and P4. 56% 45% 32% 29% 46.8% 31% 6.3% 19.6% 16.7% 12.2% 9% 2% 60% 51.9% 54.5% 32.1% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1850-1889 1890-1929 1930-1969 1970-2009 Fragrant Scented Perfumed Sweet-smelling Figure 3: P LANTS & FLOWERS over time. A Diachronic Perspective on Near-Synonymy 339 In fact, these nouns account for only 6.4% of the use of scented in P1, whereas it makes up for 26.5%, approximately four times more, in P4 (cf. Figure 4). Something similar goes on with perfumed with AESTHETICS nouns: it goes from modifying these nouns less frequently than expected in P1 and P2, to being more common in P3, although not significantly so, to finally becoming a type in P4. As seen in Figure 5, which shows the percentages of use of this adjective with AESTHETICS nouns, there is a rather drastic increase from P3 onwards, going from 2.6% in P1 to 18.4% in P4. Some important changes are going on as concerns the collocational patterning of the adjectives and, what is more, they are particularly interesting given that the decreases and increases take place in categories containing examples of the adjectives used exclusively in one of the senses: P LANTS &F LOWERS , including sense 1 uses only, and C LEANING and A ESTHETICS , including sense 2 uses only. This could point to a decrease of the adjectives in sense 1, and an increase in sense 2. Nonetheless, this interpretation is rather tentative due to lack of more significant results of the individual changes of each of the four synonyms. Therefore, to further investigate this issue, the collocational patterns Table 7: C LEANING over time. Adj. Period Fragrant Perfumed Scented Sweet-smelling Total P:–  A A P:–  A    P:–  A T T P:–  A  TT Total:      Table 8: A ESTHETICS over time. Adj. Period Fragrant Perfumed Scented Sweet-smelling Total P:–      A P:–  A  P:–     T T P:–   TT Total:  A T   340 Daniela Pettersson-Traba 2.6% 2.4% 14.6% 18.4% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% P1 P2 P3 P4 Figure 5: Perfumed with A ESTHETICS over time. 7% 14.6% 22.9% 20.4% 6.5% 14.9% 32.7% 26.5% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% P1 P2 P3 P4 perfumed scented Figure 4: Perfumed and scented with CLEANING over time. A Diachronic Perspective on Near-Synonymy 341 of the whole concept SWEET - SMELLING , as designated by all four adjectives, were analyzed. A further HCFA test was carried out with the two independent variables “semantic category”and “period”, since this configuration permits to identify the collocational patterning of the concept with the different categories across the four periods. This last set of results reveals the same tendencies for the whole concept SWEET - SMELLING as those obtained for the individual changes of each synonym, namely that PLANTS & FLOWERS goes from being an antitype to becoming a type, and that both CLEANING and AESTHETICS nouns go from being antitypes to becoming types (cf. Tables 6, 7, and 8). Additionally, another significant result appears concerning the category BODY . These nouns go from being modified less frequently than expected by the concept (although not significantly so) in the first two periods to becoming more frequent in the last two periods, being a type in P3. In the remaining semantic categories there are no statistically significant results. However, there are cases in which there exists a change in the relation between observed and expected frequency (see Section 3.2), that is, in which some semantic categories go from being either more frequent than expected to less frequent with the concept, or vice versa. For instance, nouns belonging to EARTH ,oneof the categories in which the adjectives are used only in sense 1, occur more frequently than expected in P1 and P2, but less frequently in P3 and P4. This is also the case of SENSATION nouns, which are more frequent than expected in P1, but less frequent in the three remaining periods. The reverse holds for nouns belonging to MATTER , since they go from being less frequent than expected in P1 to becoming more frequent in the rest of the periods (cf. Figure 6). The results discussed in the last paragraph seem to strengthen the interpretation put forward before: whereas the use of the adjectives in sense 1 decreases substantially, their use in sense 2 increases considerably. However, an analysis with a new variable, “sense”, would be beneficial and probably necessary to further reinforce this claim. This is so because, as mentioned in Section 3.2, it has been proven that the three semantic categories of nouns SENSATION , MATTER , and BODY include occurrences of the adjectives in both sense 1 and 2. Thus, this further variable would make it possible to tease apart the two senses and determine whether there is in fact a decrease and increase of senses 1 and 2, respectively. 4.4 General discussion of the results The present discussion focuses on three different sets of results: (1) the concept SWEET - SMELLING , (2) the relationship between the members of the synonym set, 342 Daniela Pettersson-Traba and (3) the individual development in collocational behavior of each of the four adjectives. First, as concerns the concept, there is a general decrease in frequency over 1850–2009, which is due mostly to the drastic decline of fragrant, Figure 6: Changes in the distribution of the concept of SWEET - SMELLING across the semantic categories throughout 1850–2009. A Diachronic Perspective on Near-Synonymy 343