All you need is love: metaphors of love in 1946–2016 Billboard year-end number-one songs

This study examines the use of metaphors, metonymies and metaphorical similes for love in a corpus of 52 year-end number one hit songs in theUSA from 1946 to 2016 according to Billboard charts. The analysis is performed within the framework of Conceptual Metaphor Theory and from quantitative and qualitative perspectives. Our findings indicate that the theme of romantic love is prevalent in USmainstream popmusic over the course of seven decades but shows evolutionary features. Metaphors of love evolve from conventional to novel with a notable increase in both heartbreak and erotic metaphors. Remarkably, the study finds that the two predominant conceptualizations of love in pop songs –which in a significant number of cases overlap – are the following: experiential, originating in the physical proximity of the lovers, and cultural, reflecting possession by one lover and showing a non-egalitarian type of love.


Introduction
The central theme of a large number of pop songs is some facet of romantic love. Starr and Waterman (2003: 105-110, 199-200) noted this to be already the case in the Tin Pan Alley era in the USA of the 1920s and 1930s and the trend continued through the 1940s and 1950s, when the entertainment industry grew exponentially: "total annual record sales in the United States rose from $191 million in 1951 to $514 million in 1959" (Starr and Waterman 2003: 252). In this paper, we refer to pop songs as those designed by the music industry with the intention of appealing to a large audience and, as well, to a product well-liked by many people (see Werner (2018) for a general discussion of the concept of pop culture). Indeed, pop music is created with the aim of achieving commercial success, and the evocation of feelings of falling in (and out of) love unfailingly exerts a powerful attraction on all kinds of audiences. However, as Machin (2010: 11) has suggested, "it is clear that love songs have changed over the years and can, therefore, be expected to reveal broader cultural changes in terms of gender-and sexuality-related discourses".
The language of pop culture has been largely overlooked in linguistics, and although the number of empirical studies has grown in recent times, the discourse of pop song lyrics is still very seldom analysed, especially from the perspective of metaphors. Within the framework of cognitive linguistics, Conceptual Metaphor Theory (CMT) posits that metaphors in texts, whether written or spoken, provide access to conceptualizations by their authors. Furthermore, metaphors play an important role in understanding the experience and meaning of emotions like love, chiefly because "a large part of our emotional understanding seems to be based on metaphor" (Kövecses 1990: 46). The goal of this research is, for the most commercially successful US song hits from 1946 to 2016, to identify the metaphors used to conceptualize love, the associated feelings and experiences, the different visions of love they configure, and how pop song love metaphors have evolved over time.
This paper is divided into four further sections. Section 2 reviews the literature regarding CMT and linguistic analyses of pop music lyrics; Section 3 describes the corpus under study and methodology; Section 4 describes the different perspectives offered by quantitative (including diachronic) and qualitative analyses; and, finally, Section 5 presents our conclusions.
2 Literature review CMT, a major area of interest within the field of cognitive linguistics, was initially formulated by  and subsequently developed in numerous publications, most notably in seminal works by Johnson (1987), Lakoff (1987), Lakoff and Turner (1989), Lakoff (1993) and Gibbs (1994). In their groundbreaking work, Lakoff and Johnson (1980: 3) argue that metaphors are pervasive in our day-to-day lives, not only in language but also in thoughts and actions: "Our ordinary conceptual system, in terms of which we both think and act, is fundamentally metaphorical in nature". Lakoff (1993) further argues that conceptual metaphor is the chief mechanism for understanding abstract concepts and performing abstract reasoning.
Within CMT, conceptual metaphor (CM) is defined as a cognitive process of mapping between conceptual-semantic domains, such that certain properties or structures of a source domain (SD), which is more concrete, closer to experience or better understood by the speaker, are transferred to a target domain (TD) that is usually more abstract or intangible. Accordingly, the chief function of a CM is to facilitate understanding and communicative efficacy with regard to the TD. An example is the use of food expressions to represent ideas (1).
(1) CM: IDEAS ARE FOOD ME: a. That's food for thought b. I just can't swallow that claim c. His idea was half-baked It should be noted that a single CM may be linguistically performed in multiple metaphorical expressions (ME). Put another way, MEs are surface implementations or specific linguistic cases of a CM. CMs are conventionally coded using the formula 'TARGET DOMAIN' IS 'SOURCE DOMAIN'. It is important to note the difference between conventional and novel or creative CMs. Conventional CMs "structure the ordinary conceptual system of our culture" (Lakoff and Johnson 1980: 139) and are so embedded in everyday use of the language that their metaphorical nature usually goes unnoticedas exemplified in (1). Novel CMs provide a new way of understanding our experiences, as they imaginatively establish "some novel mapping between different conceptual domains" (Gibbs 1994: 262). For their part, Bowdle and Gentner (2005) propose an evolutionary path ("the career of metaphor") which postulates a shift in mental processing from comparison to categorization as novel metaphors evolve to conventionalized metaphors. Previously, Johnson (2003[1980]: 248) have suggested that the repeated use of metaphors leads to their conventionalization, and this, in turn, often results in their lexicalization: "the conceptual metaphor explains the systematicity of the polysemy, and correspondingly, the systematic polysemy provides evidence for the existence of the metaphor". Therefore, lexicalization, i.e. setting in the lexicon, is the maximum degree of conventionalization of a CM.
Another key notion in CMT concerns how conceptualizations involve hiding and highlighting. According to Lakoff and Johnson (1980: 10), "the very systematicity that allows us to comprehend one aspect of a concept in terms of another (…) will necessarily hide other aspects of the concept". Hence, metaphors may impose a perspective onto a concept or a conceptual domain such that the use of different types of CMs to refer to a single domain may reveal different conceptions of that domain. This assumption is the grounding of the use of CMT to detect ideas or assumptions implicit in discourses.
CMs for love have been thoroughly discussed in the cognitive linguistics literature since the publication of the seminal work by , but especially in the analysis of emotions (e.g. Kövecses 1990Kövecses , 2005Kövecses , 2014 in English and specific languages and cultures (e.g. Gatambuki 2014). Text analysis from a CMT perspective has been applied to the study of poetry and literature, giving rise to the field of cognitive stylistics (e.g. Turner 1989 andSemino andCulpeper 2002). As for specific analyses of CMs for love, particularly noteworthy is Barcelona's (1995) work on Shakespeare's Romeo and Juliet.
As for application of CMT to the analysis of pop and rock songs, the few studies that do exist include Steen's (2002) in-depth empirical analysis of Bob Dylan's song Hurricane from the point of view of cognitive stylistics, and Kennedy's (2013) detailed exploration of metaphors and myths in psychedelic rock lyrics and their influence on the cultural revolution of the 1960s. We are unaware of any specific study of CMs for love in hit pop songs, other than Kreyer's (2012) analysis of variation and creativity in the lyrics of the 48 top-selling US albums of 2003. Kreyer, who analysed sentences containing the word "love" (and its variant "luv") and indicated that he identified the metaphors intuitively, proposed using the metaphor identification procedure (MIP) developed by the Pragglejaz Group (2007) to achieve greater methodological consistency in future studies.
The lyrics of hit pop songs, however, have been analysed from other linguistic perspectives, usually using corpus linguistics for the purpose of studying their status as a register. For instance, Bértoli-Dutra (2014), Motschenbacher (2016) and Brett and Pinna (2019) examine lexicogrammatical and semantic features; Trotta (2013) explores the use of non-standard grammar; and Jansen and Westphal (2017) examine morphosyntax and phonology. Both Hilbert (2012) and Kreyer and Mukherjee (2007) analyse pop song style, with the latter also exploring linguistic metaphors for the words "heart" and "love". Song lyrics have also been analysed from a sociolinguistic perspective (Coupland 2011;Squires 2019). As far as diachronic studies are concerned, Werner (2012) has explored lexicogrammatical and morphosyntactic characteristics in two corpora of UK hits (1,128 number-one hit singles from 1952 to 2008, according to the UK's Official Charts Company 1 ) and US hits (the top five for each year between 1946 and 2005, according to Billboard's year-end charts 2 ), finding that innovative patterns emerged particularly in the period from 1991 to 2005, while several conservative features were simultaneously preserved. These and other findings by Werner (2012) contribute importantly to establishing the global stylistic patterns of pop-music lyrics. From the point of view of social psychology, Pettijohn and Sacco (2009) have analysed number-one songs in the annual rankings of Billboard's year-end charts from 1955 to 2000, finding a correlation between socioeconomic changes in the USA and both song structures and themes and the physical appearance of the artists.
The core nature of the concept of love in pop music has been tested in quantitative lexicogrammatical works. Several studies have verified that "love" is the most frequent content word in the corpora of song lyrics (Werner 2012;Motschenbacher 2016). Dukes et al. (2003) have conducted a diachronic analysis of the lyrics of the 100 top-selling songs in the USA (according to Billboard magazine) between 1958 and 1998, clustered into the periods 1958-1972, 1976-1984 and 1991-1998. The methodology is based on inductive coding and the frequency of the words representing expressions of love and sex. The quantitative findings have shown that the percentage of words representing love fell over time, whereas the percentage of words representing sex peaked in the period 1976-1984, with the authors proposing sociological race-and gender-related hypotheses as possible explanations. Regarding sex, pop song lyrics have also been studied by researchers interested in language and sexuality (e.g. Motschenbacher (2018) from a Critical Discourse Studies perspective).
As can be observed, no previous study has investigated, within the framework of CMT, the overall conceptualisation of love in hit pop songs and how this conceptualization has evolved over time. Our research therefore provides an opportunity to advance in knowledge regarding what CMs reveal about this allegedly crucial component of the lyrics of the best-selling singles.

Data and method
In this section we describe the constitution of the corpus under study, followed by the analytical procedure used for detecting and formulating CMs.

Corpus constitution
To build the corpus to be analysed, we collected the lyrics of number-one hit songs, as listed in Billboard's year-end charts in the pop category, from 1946 (the first listing) to 2016 (when this study commenced). The lyrics were downloaded from Internet repositories and were checked by close listening to each song to create a transcription. The Billboard year-end charts, which name the top single and top album in the USA each year, are compiled from the weekly Billboard Hot 100, 3 the US music industry's standard records chart (published by Billboard magazine, and, more recently, on the Billboard website). Until 1991, the Billboard Hot 100 charts were calculated from reported record shop sales, airplay and, initially, jukebox activity. In 1991, Billboard began using data from Nielsen SoundScan, which tracks a large amount of actual physical sales and, more recently, online services, with the result that calculations turned out to be far more accurate (Campbell 2006:18). The corpus of 71 songs can therefore be considered to be representative of the most popular music in the USA, with sales in the millions and the greatest possible public impact as popular music that is "mass-reproduced and disseminated via the mass media [and] that has various times been listened to by large numbers of Americans" (Starr and Waterman 2003: 6). Consequently, it can also be assumed that these songs' lyrics are very likely to reflect concepts that, at any given time, were generally accepted by the society that consumed them.
Given that our focus is on the public impact of lyrics, in our final list (see Appendix) the number one was replaced by the number two in four of the 71 songs in two circumstances: when the number one was an instrumental song and when the song was in a language that was not English (as meaning would not have been fully understood by most listeners). Next, the theme of these songs was established, and the songs not about love were eliminated, leaving us with a corpus of 52 songs. This categorization was carried out by one co-author and was verified by the second co-author to ensure accuracy and consistency.
Lastly, given the great amount of textual repetition in the pop song genre, the corpus was normalized: we eliminated song titles, repeated choruses, minor variations (basically phonetic additions that add no semantic content), repeated verses and verses consisting of onomatopoeias. 4 The resulting corpus consists of a total of 8,980 words.

Analytical procedure
Two annotators, experts in Conceptual Metaphor Theory (CMT), independently identified and classified all the metaphorical expressions (MEs) in the corpus, and inferred and formulated the corresponding conceptual metaphors (CMs). Initial agreement was 71.75% and complete consensus was attained after a number of discussion meetings.
To detect MEs, the standard Metaphor Identification Procedure (MIP) (Pragglejaz Group 2007) was employed. The use of this methodology makes it possible to detect either creative or conventionalized metaphors with an active metaphorical basis (p. 30). For metaphor formulation, an adaptation of Steen's (1999Steen's ( , 2007 methodology was used. The consecutive application of these two methods was formalized into the following five steps: 1. Identification of the metaphorical focus. The word or group of words leading to the identification of the ME is detected. To carry out this step, the criteria defined by the MIP are followed: a) establishment of the meaning of the lexical unit(s) within the analysed context; b) consultation of a reference dictionary to determine whether a more basic contemporary meaning exists (more concrete, easier to imagine or perceive by the senses, more precise, related to bodily action); and c) decision as to whether the contextual meaning contrasts with the basic meaning but can be understood in comparison with it. 2. Identification of the Target Domain (TD) of the CM, based on the ME or its context. 3. Identification of the Source Domain (SD) of the CM, based on the metaphorical focus. 4. Establishment of the CM's ontological correspondences (i.e. mapping of entities between SD and TD) and epistemic correspondences (i.e. which knowledge of the SD is mapped onto the TD). 5. Detection of the level of generalization in the CM, i.e. whether the CM identified is a particular form of a more generic CM or whether it could be the combination of other more basic metaphors.
To establish the meaning of the lexical units and apply the MIP, we chose Collins English Dictionary as the reference dictionary for three main reasons: its prestige, its online availability and, above all, the fact that it is corpus-based and so provides real usage examples. For doubtful or complex cases, we used the Oxford English Dictionary as a second reference. Only CMs associated with love were sought and accounted for. The criterion used was to select CMs with a TD that was directly related to love (e.g. LOVE, THE OBJECT OF LOVE, INTIMATE SEXUAL BEHAVIOUR) or had love-related concepts as a subtype (e.g. EMOTION).
To identify CM domains, we compiled and systematically consulted a compendium of metaphors of love as detected by Kövecses (1990Kövecses ( , 2014, Barcelona (1995) and Gatambuki (2014), and metaphors of love and emotions contained in the Master Metaphor List (Lakoff et al. 1991), a general-purpose repository. For the purposes of our analysis, the CMs identified in the above-mentioned studies were regarded as conventional CMs given that the experts posit that they are well-established. Accordingly, such conventional CMs were formulated using the same labels (SD and TD) as in the reference literature, whereas novel CMs were formulated using the same criteria as those used by the above-mentioned authors.
At this point, it is worth mentioning that the Master Metaphor List declares itself to be a compilation taken from published works since those of Reddy (1979) and . But, more crucially, it has to be noted that the authors, working within the framework of CMT, claim that "our ordinary conceptual system (…) is fundamentally metaphorical in nature" (Lakoff and Johnson 1980: 3). Accordingly, the many examples of CM that the authors offer in their works are assumed to be deeply entrenched in human thought, so that they tend to be resistant to change over time. More specifically, Kövecses (1990: 4) claims that CMs "produce certain prototypical cognitive models associated with particular emotions", e.g. love, and, correspondingly, he presents a prototypical cognitive model of love consisting of most of the CMs included in our compendia, stating further that the material on which the study is based "is composed of those linguistic expressions that are commonly used by and are familiar to most, if not all, native speakers of English" (pp. 43-44). Interestingly, most of Barcelona's (1995) love metaphors retrieved from Shakespeare's Romeo and Juliet (circa 1591) are deemed to be conventional in 20th century CMT research (e.g. LOVE IS UNITY, LOVE IS A JOURNEY, THE HEART IS THE CONTAINER OF EMOTIONS).
Thus, the metaphors that we label as conventional have been existing long before the compendia were compiled and still endure today. Note that it is important not to confuse novel CMs with new linguistic expressions of conventional metaphors; in our work, the latter are counted within the conventional metaphors.
The current study takes into account not only CMs but also metaphorical similes (2a) and conceptual metonymies (2b); and image metaphors (2c) are singled out from amongst the CMs. Metaphorical similes (conceptual figures that contain comparison connectors, usually "is like") were taken into account because, as Bredin (1998: 74) indicates, they have a discursive function similar to that of CMs in that they "assert or deny a likeness between two things in such a way that one of them describes the other" and, as in conceptual metaphors, the entities compared belong to different domains. Conceptual metonymies were also included since, like metaphors, they allow to understand one concept in terms of another. Nevertheless, in this case, the relationship between them is one of causal or physical contiguity within the same conceptual domain (Lakoff and Turner 1989: 100-104), e.g. in (2b) below laying stands for making love. Finally, image metaphors are differentiated from other CMs because, as Lakoff (1993: 227) points out, instead of mapping one complex conceptual domain onto another, "image metaphors (…) are 'one-shot' metaphors: they map only one image onto another image".
(2) a. [1957] Her lips are like a volcano when it's hot b.
[1972] And the first time ever I lay with you c.
[1972] I thought the sun rose in your eyes To establish song themes according to different facets of the concept of romantic love, we used Oxford English Dictionary definitions for the headwords romance, heartbreak, sensuality and sex (Table 1). This categorization was used to assess if the diachronic distribution of conceptual figures was consistent with the general thematic distribution over time.

Analysis
This study encompasses a period of 71 years, from 1946 to 2016. As noted above, of the 71 number ones, 52 are songs whose theme is love (73.23% of the total). Hence, the prevalence of this theme in the initial corpus is extraordinarily high. This distribution confirms the continuing predominance of the model established by the Tin Pan Alley song factory and the subsequent massive deployment of the record industry (Starr and Waterman 2003: 105, 199). While the trend has remained unchanged over the course of the 71 years covered by the study, nuances and evolution were evident, as shall be revealed, in the type of CM used to evoke love and the thematic subtypes. This section examines the question from three different perspectives: Section 4.1 provides a global quantitative analysis of the distribution and predominant CMs in the corpus; in Section 4.2 a diachronic quantitative analysis looks into the evolution of conventional and novel metaphors and their related themes; Section 4.3 briefly presents the thematic distribution of the corpus; and finally, in Section  Table 2 shows the total amount of conceptual figures for love detected in the corpus. In total, there are 393 ME (tokens) corresponding to 102 different types. As can be seen, the dominant type of trope is, by far, the conceptual metaphor (71.56%), and even more so in terms of number of tokens (88.04%), because the ratio of ME per CM (4.73) is considerably higher than for metaphorical similes, conceptual metonymies and image metaphors. Turning to the relative density of conceptual figures for love in the corpus, the 52 songs consisted of 8,980 words distributed in 1,272 verses, reflecting an average of 1.96 CMs and 7.55 MEs per song, 5 with a frequency of appearance of one ME every 22.84 words, and one ME every 3.23 verses. A hit love song therefore typically includes two CMs performed in eight MEs. Table 3 shows the conceptual figures with the greatest presence in the corpus, instantiated in 10 or more MEs. All are metaphors, reflecting the quantitative preponderance of metaphors over similes, metonymies and image metaphors.

Global quantitative analysis
All the predominant CMs for love in the corpus were previously formulated in the literature as conventional (see Section 3.2). Notable examples of the 10 predominant types of CMs are reproduced in (3). Years in square brackets identify songs, in this case, the song that includes an ME (see Appendix).
(   One interesting case is LOVE IS A PLACE, as it has not been explicitly postulated in the reference literature as a conventional metaphor. However, it must be interpreted as a specific-level metaphor of the well-established generic-level metaphor STATES ARE LOCATIONS (Lakoff 1993). According to Lakoff (1993: 18), metaphorical mappings "are sometimes organized in hierarchical structures, in which 'lower' mappings in the hierarchy inherit the structures of the 'higher' mappings". Moreover, note that the highly conventional metaphor THE OBJECT OF LOVE IS A SMALL CHILD, even though it appears to be one of the most frequent CMs in our corpus, is actually instantiated in most cases with the same ME, i.e. the very common expression "baby" to refer to the loved one (25 of the 38 occurrences of this CM).
Finally, note also that there is not a single love song that does not contain at least one metaphor for love, while just three ([1951], [1961] and [1974]) feature only one. Figure 1 shows distributions over time of the conventional and novel metaphors for each decade represented in the corpus. 6 As mentioned, in this research those metaphors recorded in the specialized literature are deemed conventional. The period reflecting the greatest presence of metaphors of love in the corpus (conventional or novel) runs from the mid-1970s to the mid-2000s, particularly from 1996 to 2005, which accounts for 26% of the corpus metaphors. As for novel metaphors, the period from the mid-1970s to 2016 accounts for 85% of all novel metaphors, and particularly the 1996-2005 period.

Diachronic quantitative analysis
With regard to first appearances of a type of metaphor in the corpus, Figure 2 shows how the study period splits naturally into three sub-periods: 1946 to the end of the 1950s, with many first appearances of conventional metaphors; the end of the 1960s to the beginning of the 1990s, with comparable numbers of both conventional and novel metaphors appearing for the first time; and the period from the end of the 1990s, marked by a preponderance of first appearances of novel metaphors.
Interestingly, metaphors appearing in the early years tend to endure: the vast majority of those appearing for the first time up to the early 1970s (except for three) reappear subsequently, and many endure to the present day. The most enduring metaphor (70 years, 1947 to 2016) is LOVE IS CLOSENESS. In contrast, most CMsmostly novelappearing from the end of the 1970s appear only once. More  Metaphors of love in number-one songs specifically, 35 of the 40 novel metaphors identified in the corpus appear just once, pointing to their non-conventionalization. Exceptionally, of the metaphors appearing in the corpus from the late 1970s, four turn out to be stable, with more than five appearances each over more than 20 years; three of these are conventional, specifically, LOVE IS A VALUABLE ASSET (4a), LOVE IS WAR (4b) and LOVE IS A JOURNEY (see 3j above), and the fourth is the novel LOVE IS A GAME (4c).
[2011] You'll pay me back in kind and reap just what you sow. b.
[1991] Yeah, I would fight for you. c.
[1992] I have no time for you to be playing with my heart like.
It is significant that almost half (15 of 35; 43%) of the novel metaphors appearing subsequent to 1976 are related to the domains of sensuality (5a, 5b) and sex (5c, 5d), in contrast with only a fifth (1 of 5; 20%) in the period before 1976.
(5) a. [1980] Cover me with love b. [1982] Let me hear your body talk. c. [1980] Call me for a ride.

d. [2004] These women all on the prowl
Overall, these results would suggest that the years up to the mid-1970s were characterized by the appearance in the corpus of highly conventionalized metaphors, whilst the years thereafter are characterized by the continuing use of previously used conventional metaphors, the large-scale creation of novel metaphors that appear only once or are rarely repeated, and a significant increase in novel metaphors in the sensuality and sex domains.

Thematic distribution
Not surprisingly, the diachronic distribution of metaphors is consistent with the thematic distribution over time. Classification by theme subtypes resulted in a corpus naturally split into two main congruent blocks, one romantic (romance and heartbreak) and the other erotic (sensuality and sex), consisting of 39 songs (75%) and 13 songs (25%), respectively. Therefore, the majority of love songs in the corpus are romantic. Songs with an erotic theme are concentrated in 1977-1982 (6 songs in 6 years) and, to a lesser degree, in 1996-2008 (5 songs in 13 years). Of the romantically themed songs, almost all of those reflecting heartbreak (14 of 15) were concentrated in the period from 1983 to 2016, with just one occurring in an earlier period (specifically in 1956).
While space limitations do not permit further analysis, it is remarkable to note that songs with a romantic theme (romance and heartbreak) are in the majority, especially in the earlier period reflected in the corpus (to the end of the 1970s). In the first part of the corpus the main theme is romance, while heartbreak also becomes commonplace from the 1980s. At the end of the 1970s and in the early 1980s there is a brief preponderance of erotically themed songs, which also had an impact at the turn of the century.

Qualitative analysis: predominant metaphorical structure in love songs
As was pointed out in Section 2, in a conceptual metaphor a TD (the domain to be understood) is mapped from a SD (which, by comparison, enables understanding of the TD). Hence, the TD determines the themes dealt with and the SD determines how they are conceptualized. Firstly, we note two characteristics specific to the conceptual figures other than metaphors in the corpus. Metonymies typically have physical effects as explanatory concepts of love, e.g. INABILITY TO BREATHE or THINK (6a, 6b) or INCREASE OF BODY HEAT (6c). As for the image metaphors in our corpus, these are all erotic (7a, 7b), whereas the similes do not display characteristics different from those of conceptual metaphors.
(7) a. [1977] Spread your wings and let me come inside b.
[1998] Baby when we're grinding With regard to the predominant conceptual figure (the conceptual metaphor), it comes as no surprise that by far the most common TD in the corpus in terms of number of occurrences is LOVE, with 216 MEs, followed at a distance by THE OBJECT OF LOVE (57 MEs) and the HEART (22 MEs). This would indicate that hit pop songs use love-as-a-concept metaphors four times more than they use personloved metaphors. THE OBJECT OF LOVE is mostly explained in terms of a DEITY (8), SMALL CHILD (3b) or APPETIZING FOOD (3f). The HEART in all cases is dealt with using the common cliché THE HEART IS THE CONTAINER OF LOVE (3d).
(8) [1946] From one who's master of my fate More interesting is the composition of the SDs, as the determinants of the metaphorical conceptualizations. Their makeup, as might be expected, is much more diverse than that of the TDs (which are necessarily concentrated in the most important concepts to be explained or understood). However, amongst this great dispersion, two dominant conceptualizations stand out (see Table 3). First, and above all, it should be noted that the concepts of CLOSENESS (9a, 3a), PHYSICAL CONTACT (9b, 3g) and UNITY (9c, 3e) taken together total 70 MEs in 29 different songs.
[2000] I'm closer than I've ever felt before b.
[1964] I wanna hold your hand c. [1968] Remember to let her into your heart Kövecses (1990Kövecses ( : 132, 2005 picks up Lakoff's (1987) notion that metaphorical thought is largely based on correlated experience: one of the best examples is the correlation between intimacy and physical closeness. Kövecses (2014: 26) further states that emotion concepts are largely metaphorical and metonymic in nature: "emotion metonymies serve as the basis for some emotion metaphors; that is, the metaphors emerge through a metonymic stage" since "there are important conceptual and bodily connections between some of the emotional responses and the emotion concepts they characterize". The relationship between PHYSICAL CLOSENESS and LOVE is, according to Kövecses (2014: 20), also one of the outstanding cases of emotion concepts with a bodily basis. Our findings are thus consistent with Kövecses' tenets, since, in our corpus, love is predominantly conceived of in terms of physical proximity to the lover (a situation commonly experienced or desired by them). Moreover, our findings suggest that this conceptualization is structured in three progressively closer degrees: closeness, contact, and fusion (or unity). The second predominant cluster which is drawn in quantitative terms in the corpus is that of submission: the POSSESSION (10a, 3c) and BONDAGE (10b, 3i) SDs occur 40 times in 21 different songs. This type of metaphors suggests an unequal kind of love, with one of the lovers regarding the other as "his/her own", with a number of cases pointing to this possession as constraining (a form of subjugation). In this respect, account has to be taken of the evidence that not all conceptual metaphors have an experiential origin: Kövecses (2005: 4, 7) posits that "metaphors are not necessarily based on bodily experiencemany are based on cultural considerations. (…) Conceptual metaphors converge on and often produce cultural models that operate in thought".
(10) a. [1958] I can make you mine b. [1975] Ain't gonna set you free now The physical proximity cluster is, as suggested above, metonymic or embodied in nature. However, the possession cluster is not, as there is no basic physical experience that links loving someone with possessing them or holding them in subjugation; hence, this cluster has to be inferred as a cultural model. We therefore posit that, of the two broad models for the conceptualization of love as reflected in the corpus, one is an experiential frame, arising from physical proximity, whilst the other is a cultural frame, based on the concept of possession. It should also be noted that POSSESSION and BONDAGE entail proximity (precisely the concept framed by the predominant cluster of metaphors), but the symmetrically opposite affirmation is not so common, as reflected in our corpus: 70% of the songs with metaphors of POSSESSION or BONDAGE also incorporate metaphors of CLOSENESS, PHYSICAL CONTACT or UNITY, and, in contrast, only 48.28% of songs with metaphors of physical proximity feature metaphors of possession. Lastly, it should be emphasised that 67.31% of the songs in the corpus include metaphors from either of these two predominant clusters: physical proximity or possession.
To sum up, our corpus analysis provides evidence that the two main conceptualizations of love in the top US hit pop songs released in the period 1946-2016 are physical proximity (experiential) and possession (cultural), and also that, in a significant number of cases, the former is implied by the latter in such a way that both frames occur jointly.

Conclusions
To our knowledge, this is the first study to perform, within the framework of CMT, a methodologically sound and fully manual (i.e. not based on keyword extraction) analysis of a diachronic corpus representative of the most commercially successful songs in the USA from 1946 to 2016, with the aim of detecting the predominant conceptual metaphors of love in their lyrics. Songs about love in any of its varied facets represented almost three quarters of the year-end number-one hit songs for the period.
In total, we detected 393 metaphorical expressions corresponding to 102 conceptual metaphors, conceptual metonymies or metaphoric similes. Number one hit songs typically contained an average of eight metaphorical expressions corresponding to two types of conceptual metaphors, and all the love songs used at least one metaphor for love.
The prevalence of the theme of romantic love in mainstream pop music in the USA over the course of seven decades is confirmed in a corpus of the best-selling pop songs which, moreover, suggests several evolutionary features in the type of conceptual metaphors used andrelatedlyin the theme of the songs. The distribution in the corpus progresses from an initial period of conventional metaphors that are repeated over the course of time to the progressive emergence (from the mid-1970s on, and particularly between 1996 and 2005) of novel metaphors that were rarely or never repeated and the first appearance of numerous eroticallythemed metaphors.
Our corpus analysis of metaphors revealed two predominant conceptualizations of love: one experiential, originating in the physical proximity of lovers, and the other cultural, based on the concept of possession of one lover by the other. In the former, love is conceptualized in terms of three progressive degrees of proximity: LOVE IS CLOSENESS, LOVE IS PHYSICAL CONTACT and LOVE IS UNITY. The cultural frame reflects a non-egalitarian type of love evidenced by the metaphors LOVE IS POSSESSION and LOVE IS BONDAGE. In a significant number of songs, the two frames overlap.
To conclude this study of the conceptual metaphors of a diachronic corpus of year-end number one hit songs in the USA, we would like to suggest other themes and directions which could enlarge our findings and enrich the discussion. Given that manual annotation of a relevant corpus has been useful and necessary in identifying prevailing CMs, the study could be further extended, for example to top-10 songs, to see whether the observed trends are confirmed. To tackle this larger corpus, semiautomatic analytical strategies could be used, e.g. those proposed by Stefanowitsch (2006). It might also be interesting to extend the study to songs from other cultures and languages so as to determine whether patterns were the same or different. Furthermore, extending the study to a corpus of more recent years could contribute to assessing whether the novel metaphors we have encountered are in the process of conventionalization. Lastly, from a more reflective point of view, our findings could be analysed from the broader perspective of language and sexuality, given that love song lyrics are a typical example of the discursive construction of sexual desire.