TY - JOUR
AU - Regan, John
AB - Abstract This article is a computational enquiry into the different ways in which two words, assumed to be central to the eighteenth-century concept of aesthetics, were used across that century. Using word co-association measures designed specifically for this study, I show the markedly different lexis that surrounded the words ‘beauty’ and ‘beautiful’ in three decades of historical textual data from Eighteenth-Century Collections Online. Having demonstrated that these words were used in very different semantic contexts in the beginning, middle, and end of the century, the article presents semantic network diagrams which further demonstrate the distinctness of noun from adjective. I argue that while ‘beautiful’ is suspended in a semantic space that is recognizably aesthetic, ‘beauty’ is not. Furthermore, that which is deemed ‘beautiful,’ according to the evidence presented here, presents something of a challenge to received narratives of how the concept of aesthetics developed throughout the eighteenth century. The objective of this article is to investigate the uses of the words ‘beauty’ and ‘beautiful’ in various historical slices of the eighteenth-century corpus. It employs a custom-designed, computational measure of word co-association to explore the semantic fields within which these words were used in three sections of Eighteenth-Century Collections Online (hereafter ECCO).1 Findings will be presented numerically at first, and then in the form of diagrammatic semantic networks. These diagrams show not only how the semantic contents of these words' lexical surroundings altered through time, but how structured relations between words were also susceptible of dramatic historical shifts. In uncovering these features of these words' uses across the aggregated whole of the corpus, this work opens windows onto the most common ways in which these words were used in different historical periods of eighteenth-century print.2 In order that the results and findings here will seem purposeful to readers as they read, it seems warranted to articulate this work's two main findings immediately, so that the numbers are diagrams will be inspected to some purpose. The headline claims are as follows. A view onto the aggregated whole of ECCO in different historical periods tells us that ‘beauty’ and ‘beautiful’ were employed (again, in the common, aggregated view across the corpus of ECCO) in markedly distinct semantic domains. Throughout the century, ‘beauty’ is most commonly situated in semantic networks that resemble constructions of cosmetic femininity. By contrast, ‘beautiful’ is, in this common view into what was printed, at the center of semantic networks of artistic craft and materiality: the language of sculpture, architecture and landscaping. The semantic surroundings of the adjective broadly, with important and interesting qualifications, align with those around it in eighteenth-century aesthetics. Finally, by way of introduction, it is important to make clear the nature of the claim that will be made here, relative to what we know about the history of British aesthetics in the eighteenth century. In succinct terms, this is intended as a complementary, rather than contradictory, story of the semantic lives of two words in the eighteenth-century corpus to what currently obtains in aesthetics. The aim is decidedly not to assert that, because the common stock of printed materials in given historical periods reveals perhaps surprising semantic environments around these two words, that the history of the concept of aesthetics must be rethought. Rather, it is thought that these views onto the life of these terms in their most common semantic fields can provide fascinating complements to their more specialized suggestiveness in, say, philosophical discussions of taste and affect. In other words, what is shown here is how the diverse printed polity used ‘beauty’ and ‘beautiful,’ with the thought that evidence of common uses can be compared in fascinating ways to the innovative uses of these words in aesthetic discourse. I. DIFFERENT, EVOLVING LEXICAL ENVIRONMENTS FOR ‘BEAUTY’ AND THE ‘BEAUTIFUL’ In all forms of linguistic expression, certain words are more likely to appear in proximity to one another, than others. The word ‘parliament’ is more likely to occur in sentences with the word ‘Thames’ than with ‘golf,’ for example. Several methods in corpus linguistics have grappled with, and in their own ways solved, the problem of accounting for the likelihood of words co-associating. We may now quantify how much more likely ‘parliament’ and ‘Thames’ are to occur than ‘parliament’ and ‘golf,’ in historical corpora. This article uses a measure that has been custom-built for this research by the author in collaboration with other computational scholars, called Distributional Probability Factor (hereafter dpf).3 As much information about how dpf works as is feasible within the space provided, will be given in the footnotes here.4 Table 1 shows the words most likely to occur at a distance of ten from the word ‘logarithm’ in the decade 1740–1750 in the ECCO corpus.5 These are in ranked order, with the most likely at the top. The top thirty words are shown here, giving a good impression of what most authors who published texts in this discursive environment would have written in sentences with this word.6 Viewing the company that words keep is important, because a large part of how meaning is constructed is through words’ relations with other words. Words which tend to occur relatively close together in lexical space produce communities in which semantic content is produced and inflected. 7 So, using the ECCO corpus as our field of enquiry, Table 1 gives a view onto the semantic environment around this word as produced by all uses of the word in every text held by ECCO, with a publication date falling between 1700 and 1750. Table 1 Words most likely to co-associate with the word ‘logarithm’ at lexical distance 10 in the 1740–1750 time slice of Eighteenth-Century Collections Online (ECCO). Note that most-strongly co-associated words appear at top of list and are listed in descending order of strength of likely co-association Words likely to occur with ‘logarithm’ at lexical distance 10. Years 1740–1750 Corpus-ECCO . logarithm 235585 logarithms 59621 log 22492 arithmetical 14171 index 13140 fraction 12798 decimal 8855 corresponding 8256 complement 8154 quotient 6811 prefix 6630 numerator 6610 unity 6263 Denominator 5337 fluent 5129 fluxion 5018 secant 4792 progression 4566 product 4499 Corresponds 4482 Words likely to occur with ‘logarithm’ at lexical distance 10. Years 1740–1750 Corpus-ECCO . logarithm 235585 logarithms 59621 log 22492 arithmetical 14171 index 13140 fraction 12798 decimal 8855 corresponding 8256 complement 8154 quotient 6811 prefix 6630 numerator 6610 unity 6263 Denominator 5337 fluent 5129 fluxion 5018 secant 4792 progression 4566 product 4499 Corresponds 4482 Open in new tab Table 1 Words most likely to co-associate with the word ‘logarithm’ at lexical distance 10 in the 1740–1750 time slice of Eighteenth-Century Collections Online (ECCO). Note that most-strongly co-associated words appear at top of list and are listed in descending order of strength of likely co-association Words likely to occur with ‘logarithm’ at lexical distance 10. Years 1740–1750 Corpus-ECCO . logarithm 235585 logarithms 59621 log 22492 arithmetical 14171 index 13140 fraction 12798 decimal 8855 corresponding 8256 complement 8154 quotient 6811 prefix 6630 numerator 6610 unity 6263 Denominator 5337 fluent 5129 fluxion 5018 secant 4792 progression 4566 product 4499 Corresponds 4482 Words likely to occur with ‘logarithm’ at lexical distance 10. Years 1740–1750 Corpus-ECCO . logarithm 235585 logarithms 59621 log 22492 arithmetical 14171 index 13140 fraction 12798 decimal 8855 corresponding 8256 complement 8154 quotient 6811 prefix 6630 numerator 6610 unity 6263 Denominator 5337 fluent 5129 fluxion 5018 secant 4792 progression 4566 product 4499 Corresponds 4482 Open in new tab In the above case of ‘logarithm,’ the language we find inflecting its meaning could, perhaps, have been predicted. This is to say that it is not particularly surprising. But when the objects of enquiry are not mathematical words surrounded by such predictable, contextual or disciplinary vocabularies, interesting things can be recovered from the historical record. ‘Beauty’ and ‘beautiful’ are two such objects of enquiry. Tables 2 and 3 contain the binding lists at distance ten for ‘beauty’ and ‘beautiful.’ Because we are interested in historical semantic change, each column shows what lexis was bound to that word in each of three decades, 1720–1730, 1750–1760 and 1790–1800. These have been chosen to give an insight into the beginning, middle and end of the eighteenth century.8 Table 2 Words most likely to co-associate with the word ‘beauty’ at lexical distance 10 in three decades of ECCO data: 1720–1730, 1740–1750 and 1790–1800 ‘Beauty’ 1720–1730, distance 10 . 1750–1760 . 1790–1800 . rebukes 4605 rebukes 4615 deformity 5019 charms 3529 charms 3818 rebukes 4943 deformity 3484 deformity 3415 beauty 4759 arrangements 3408 beauty 3080 charms 4678 beauty 3263 symmetry 2968 symmetry 3804 bloom 258 bloom 2295 bloom 2974 symmetry 2432 fade 2195 fades 2971 lovely 2341 fades 2182 lovely 2766 captivated 2252 charm 2056 blooming 2700 similarity 2239 features 1940 enamour 2612 charm 2176 lovely 1828 features 2590 features 2038 charming 1768 heraldry 2554 fade 1876 beholders 1738 softening 2536 charming 1844 fairest 1663 fade 2370 verdure 1793 blooming 1613 canker 2367 feature 1745 fair 1597 charm 2323 fading 1681 provident 1595 graces 2299 enamour 1677 cleanliness 1579 graceful 2265 blooming 1643 garment 1509 fair 2067 deform 1640 graceful 1495 complexion 2050 elegance 1625 battlements 1467 admiration 1946 surpassed 1614 fading 1460 sweetness 1925 fades 1593 bright 1451 elegance 1914 doat 1554 gardens 1449 enchanting 1866 fair 1544 unmov 1444 her 1855 helen 1509 moth 1405 venus 1849 nymph 1497 ornaments 1363 nymph 1828 dazzled 1479 attractive 1355 turbid 1819 beautiful 1404 enamour 1342 deformed 1819 her 1402 withers 1342 adorn 1813 cheeks 1396 cupid 1330 charming 1803 ‘Beauty’ 1720–1730, distance 10 . 1750–1760 . 1790–1800 . rebukes 4605 rebukes 4615 deformity 5019 charms 3529 charms 3818 rebukes 4943 deformity 3484 deformity 3415 beauty 4759 arrangements 3408 beauty 3080 charms 4678 beauty 3263 symmetry 2968 symmetry 3804 bloom 258 bloom 2295 bloom 2974 symmetry 2432 fade 2195 fades 2971 lovely 2341 fades 2182 lovely 2766 captivated 2252 charm 2056 blooming 2700 similarity 2239 features 1940 enamour 2612 charm 2176 lovely 1828 features 2590 features 2038 charming 1768 heraldry 2554 fade 1876 beholders 1738 softening 2536 charming 1844 fairest 1663 fade 2370 verdure 1793 blooming 1613 canker 2367 feature 1745 fair 1597 charm 2323 fading 1681 provident 1595 graces 2299 enamour 1677 cleanliness 1579 graceful 2265 blooming 1643 garment 1509 fair 2067 deform 1640 graceful 1495 complexion 2050 elegance 1625 battlements 1467 admiration 1946 surpassed 1614 fading 1460 sweetness 1925 fades 1593 bright 1451 elegance 1914 doat 1554 gardens 1449 enchanting 1866 fair 1544 unmov 1444 her 1855 helen 1509 moth 1405 venus 1849 nymph 1497 ornaments 1363 nymph 1828 dazzled 1479 attractive 1355 turbid 1819 beautiful 1404 enamour 1342 deformed 1819 her 1402 withers 1342 adorn 1813 cheeks 1396 cupid 1330 charming 1803 Open in new tab Table 2 Words most likely to co-associate with the word ‘beauty’ at lexical distance 10 in three decades of ECCO data: 1720–1730, 1740–1750 and 1790–1800 ‘Beauty’ 1720–1730, distance 10 . 1750–1760 . 1790–1800 . rebukes 4605 rebukes 4615 deformity 5019 charms 3529 charms 3818 rebukes 4943 deformity 3484 deformity 3415 beauty 4759 arrangements 3408 beauty 3080 charms 4678 beauty 3263 symmetry 2968 symmetry 3804 bloom 258 bloom 2295 bloom 2974 symmetry 2432 fade 2195 fades 2971 lovely 2341 fades 2182 lovely 2766 captivated 2252 charm 2056 blooming 2700 similarity 2239 features 1940 enamour 2612 charm 2176 lovely 1828 features 2590 features 2038 charming 1768 heraldry 2554 fade 1876 beholders 1738 softening 2536 charming 1844 fairest 1663 fade 2370 verdure 1793 blooming 1613 canker 2367 feature 1745 fair 1597 charm 2323 fading 1681 provident 1595 graces 2299 enamour 1677 cleanliness 1579 graceful 2265 blooming 1643 garment 1509 fair 2067 deform 1640 graceful 1495 complexion 2050 elegance 1625 battlements 1467 admiration 1946 surpassed 1614 fading 1460 sweetness 1925 fades 1593 bright 1451 elegance 1914 doat 1554 gardens 1449 enchanting 1866 fair 1544 unmov 1444 her 1855 helen 1509 moth 1405 venus 1849 nymph 1497 ornaments 1363 nymph 1828 dazzled 1479 attractive 1355 turbid 1819 beautiful 1404 enamour 1342 deformed 1819 her 1402 withers 1342 adorn 1813 cheeks 1396 cupid 1330 charming 1803 ‘Beauty’ 1720–1730, distance 10 . 1750–1760 . 1790–1800 . rebukes 4605 rebukes 4615 deformity 5019 charms 3529 charms 3818 rebukes 4943 deformity 3484 deformity 3415 beauty 4759 arrangements 3408 beauty 3080 charms 4678 beauty 3263 symmetry 2968 symmetry 3804 bloom 258 bloom 2295 bloom 2974 symmetry 2432 fade 2195 fades 2971 lovely 2341 fades 2182 lovely 2766 captivated 2252 charm 2056 blooming 2700 similarity 2239 features 1940 enamour 2612 charm 2176 lovely 1828 features 2590 features 2038 charming 1768 heraldry 2554 fade 1876 beholders 1738 softening 2536 charming 1844 fairest 1663 fade 2370 verdure 1793 blooming 1613 canker 2367 feature 1745 fair 1597 charm 2323 fading 1681 provident 1595 graces 2299 enamour 1677 cleanliness 1579 graceful 2265 blooming 1643 garment 1509 fair 2067 deform 1640 graceful 1495 complexion 2050 elegance 1625 battlements 1467 admiration 1946 surpassed 1614 fading 1460 sweetness 1925 fades 1593 bright 1451 elegance 1914 doat 1554 gardens 1449 enchanting 1866 fair 1544 unmov 1444 her 1855 helen 1509 moth 1405 venus 1849 nymph 1497 ornaments 1363 nymph 1828 dazzled 1479 attractive 1355 turbid 1819 beautiful 1404 enamour 1342 deformed 1819 her 1402 withers 1342 adorn 1813 cheeks 1396 cupid 1330 charming 1803 Open in new tab Table 3 Words most likely to co-associate with the word ‘beautiful’ at lexical distance 10 in three decades of ECCO data: 1720–1730, 1740–1750 and 1790–1800 ‘Beautiful’ 1720–1730, distance 10 . 1750–1760 . 1790–1800 . descriptive 3385 perennial 2873 elegant 2796 vivid 2420 paintings 2060 beautiful 2475 allusions 2237 granite 2017 ornamented 2458 deformed 1975 adorned 1972 variegated 2442 blossoms 1856 marble 1922 plumage 2442 adorned 1822 variegated 1911 adorned 2438 symmetry 1714 allusions 1847 scenery 2380 drawings 1655 flowering 1764 symmetry 2302 examines 1625 stalk 1716 deformed 2237 paintings 1576 gardens 1704 verdure 2179 spotted 1521 elegant 1646 delightful 2137 holly 1496 finest 1639 gardens 2063 abigail 1479 flowers 1613 finest 2060 sparkling 1453 finely 1563 magnificent 2045 beautiful 1445 attitudes 1558 interspersed 2032 stately 1443 gardening 1555 landscape 1858 lovely 1443 delightful 1535 architecture 1827 wildness 1432 symmetry 1514 tints 1813 beauty 1404 tinge 1506 ugly 1810 rainbow 1401 jasper 1500 handsome 1806 imagination 1388 beautiful 1432 paintings 1791 ranged 1385 tops 1393 beauty 1791 ornaments 1377 examines 1393 marble 1760 delightful 1375 ornaments 1346 flowers 1752 clusters 1351 feathers 1336 rainbow 1744 variegated 1325 porphyry 1304 elegance 1714 gardens 1307 charming 1221 villas 1703 allegory 1302 tinged 1212 butterfly 1654 finest 1302 cabinets 1210 charming 1646 deform 1213 ornamented 1182 hills 1626 comely 1207 painted 22331 finely 1623 ‘Beautiful’ 1720–1730, distance 10 . 1750–1760 . 1790–1800 . descriptive 3385 perennial 2873 elegant 2796 vivid 2420 paintings 2060 beautiful 2475 allusions 2237 granite 2017 ornamented 2458 deformed 1975 adorned 1972 variegated 2442 blossoms 1856 marble 1922 plumage 2442 adorned 1822 variegated 1911 adorned 2438 symmetry 1714 allusions 1847 scenery 2380 drawings 1655 flowering 1764 symmetry 2302 examines 1625 stalk 1716 deformed 2237 paintings 1576 gardens 1704 verdure 2179 spotted 1521 elegant 1646 delightful 2137 holly 1496 finest 1639 gardens 2063 abigail 1479 flowers 1613 finest 2060 sparkling 1453 finely 1563 magnificent 2045 beautiful 1445 attitudes 1558 interspersed 2032 stately 1443 gardening 1555 landscape 1858 lovely 1443 delightful 1535 architecture 1827 wildness 1432 symmetry 1514 tints 1813 beauty 1404 tinge 1506 ugly 1810 rainbow 1401 jasper 1500 handsome 1806 imagination 1388 beautiful 1432 paintings 1791 ranged 1385 tops 1393 beauty 1791 ornaments 1377 examines 1393 marble 1760 delightful 1375 ornaments 1346 flowers 1752 clusters 1351 feathers 1336 rainbow 1744 variegated 1325 porphyry 1304 elegance 1714 gardens 1307 charming 1221 villas 1703 allegory 1302 tinged 1212 butterfly 1654 finest 1302 cabinets 1210 charming 1646 deform 1213 ornamented 1182 hills 1626 comely 1207 painted 22331 finely 1623 Open in new tab Table 3 Words most likely to co-associate with the word ‘beautiful’ at lexical distance 10 in three decades of ECCO data: 1720–1730, 1740–1750 and 1790–1800 ‘Beautiful’ 1720–1730, distance 10 . 1750–1760 . 1790–1800 . descriptive 3385 perennial 2873 elegant 2796 vivid 2420 paintings 2060 beautiful 2475 allusions 2237 granite 2017 ornamented 2458 deformed 1975 adorned 1972 variegated 2442 blossoms 1856 marble 1922 plumage 2442 adorned 1822 variegated 1911 adorned 2438 symmetry 1714 allusions 1847 scenery 2380 drawings 1655 flowering 1764 symmetry 2302 examines 1625 stalk 1716 deformed 2237 paintings 1576 gardens 1704 verdure 2179 spotted 1521 elegant 1646 delightful 2137 holly 1496 finest 1639 gardens 2063 abigail 1479 flowers 1613 finest 2060 sparkling 1453 finely 1563 magnificent 2045 beautiful 1445 attitudes 1558 interspersed 2032 stately 1443 gardening 1555 landscape 1858 lovely 1443 delightful 1535 architecture 1827 wildness 1432 symmetry 1514 tints 1813 beauty 1404 tinge 1506 ugly 1810 rainbow 1401 jasper 1500 handsome 1806 imagination 1388 beautiful 1432 paintings 1791 ranged 1385 tops 1393 beauty 1791 ornaments 1377 examines 1393 marble 1760 delightful 1375 ornaments 1346 flowers 1752 clusters 1351 feathers 1336 rainbow 1744 variegated 1325 porphyry 1304 elegance 1714 gardens 1307 charming 1221 villas 1703 allegory 1302 tinged 1212 butterfly 1654 finest 1302 cabinets 1210 charming 1646 deform 1213 ornamented 1182 hills 1626 comely 1207 painted 22331 finely 1623 ‘Beautiful’ 1720–1730, distance 10 . 1750–1760 . 1790–1800 . descriptive 3385 perennial 2873 elegant 2796 vivid 2420 paintings 2060 beautiful 2475 allusions 2237 granite 2017 ornamented 2458 deformed 1975 adorned 1972 variegated 2442 blossoms 1856 marble 1922 plumage 2442 adorned 1822 variegated 1911 adorned 2438 symmetry 1714 allusions 1847 scenery 2380 drawings 1655 flowering 1764 symmetry 2302 examines 1625 stalk 1716 deformed 2237 paintings 1576 gardens 1704 verdure 2179 spotted 1521 elegant 1646 delightful 2137 holly 1496 finest 1639 gardens 2063 abigail 1479 flowers 1613 finest 2060 sparkling 1453 finely 1563 magnificent 2045 beautiful 1445 attitudes 1558 interspersed 2032 stately 1443 gardening 1555 landscape 1858 lovely 1443 delightful 1535 architecture 1827 wildness 1432 symmetry 1514 tints 1813 beauty 1404 tinge 1506 ugly 1810 rainbow 1401 jasper 1500 handsome 1806 imagination 1388 beautiful 1432 paintings 1791 ranged 1385 tops 1393 beauty 1791 ornaments 1377 examines 1393 marble 1760 delightful 1375 ornaments 1346 flowers 1752 clusters 1351 feathers 1336 rainbow 1744 variegated 1325 porphyry 1304 elegance 1714 gardens 1307 charming 1221 villas 1703 allegory 1302 tinged 1212 butterfly 1654 finest 1302 cabinets 1210 charming 1646 deform 1213 ornamented 1182 hills 1626 comely 1207 painted 22331 finely 1623 Open in new tab Tables 2 and 3 give an impression of the changing lexical environments in which ‘beauty’ and ‘beautiful’ were suspended at different points in the eighteenth century. But the lexis that is common to all three lists is also of interest: the language which binds to these words in all three periods. The words which are common to all three time slices for ‘beauty’ are: beauty bloom blooming charm charming charms deformity enamour fade fades fair features lovely rebukes symmetry The words which are common to all three time slices for ‘beautiful’ are: adorned beautiful deformed delightful finest gardens paintings symmetry variegated These bolded words provide the historically unchanging picture, while the lexis in the tables show what bound to these two words in different historical periods. And so, with both a diachronic and historically immutable picture, we can now begin to make observations about the differing lexical environments around noun and adjective. We may begin with an observation which may seem too baldly indicative to be meaningful. This is that ‘beautiful’ is not found on any of the binding lists for ‘beauty,’ and that ‘beauty’ is only found relatively lowly ranked on the binding lists for the 1720s and 1790s. This means two things. The most obvious is that these words are simply not commonly used together in sentences in these historical tranches of ECCO. Secondly, where they do appear together, there is an asymmetry in the likelihood of readers encountering the noun in sentences containing the adjective, and vice versa. This means that in general, readers were more likely to encounter the noun ‘beauty’ if the adjective ‘beautiful’ had been used, than the other way around. Put in other terms, across the eighteenth-century corpus, having read the word ‘beauty,’ it was surprisingly unlikely that one would then encounter the word ‘beautiful.’ The two words simply did not exert much force of attraction upon one another; in fact, the weakness of their relation is notable. As we shall see below, this may be attributable to the quite distinct semantic environments around these two words across ECCO: the words commonly kept quite different semantic company. Already this is a feature of their usage that may be compared with how the two words relate to one another in aesthetic treatises and books from the time. Having made this first observation, let us now begin to explore the semantic fields within which the two words were embedded across the century. Beginning with the adjective, in all three times slices above, ‘beautiful’ is surrounded by the language of objects and artefacts. In the 1720s this includes ‘paintings’ and ‘drawings,’ ‘blossoms,’ ‘gardens,’ and ‘rainbows.’ Descriptors are suited to things rather than people: ‘stately,’ ‘ranged,’ ‘vivid,’ ‘sparkling,’ ‘spotted,’ and ‘variegated.’ This is the vocabulary of art and inanimate objects, rather than what is beautiful in the human form. There is a strong feeling of landscapes being described. In the 1750s, and through the 1790s, there is the emergence of a more markedly handmade, artistic conception of the beautiful, in addition to one relating to the environment or landscape. From the mid-century (remembering at all times that this is a view onto the word's most common associations) the adjective is suspended more clearly in architectural and sculptural discursive environments. A type of artistic materiality supervenes: ‘granite,’ ‘marble,’ ‘gilding,’ ‘porphyry.’ The first signs are that this word is embedded in the domain of the material arts; a vocabulary of tactility and texture. Already we come to some sense that the lexis around the adjective ‘beauty’ overlaps with that found in discussions of taste which encompass landscaping, architecture, sculpture, and the decorative arts. Considering ‘beautiful’ in the 1750s and 1790s, it seems clear which words are modified by the adjective: ‘drawings,’ ‘gilding,’ ‘amphitheatre,’ ‘structure,’ and ‘marble’ among others. But the question of which of its other bound terms are cognates, is somewhat less clear. The following could be near or distant synonyms for ‘beautiful’: ‘elegant,’ ‘superb,’ ‘polished,’ but those adjectives relating to ornamentation appear more ambiguous: ‘decorated,’ ‘adorned,’ ‘ornamental’ and, again, that most riven word in eighteenth-century aesthetics, ‘polished.’ This word was a site of great contest throughout the proliferation of aesthetic treatises in the century. In poetics alone, across the approximately 800 publications on versification, variously used ‘polished’ to describe verse which was overly highly wrought, too metrically neat, or which relied too heavily on classical allusion and personification.9 But this was the view from the late century when neoclassicism, Alexander Pope, John Gay, and the long shadow of Dryden, were long gone. While it would be incorrect to assert that these poets and theorists believed that “polished” meant “good” verse, they undoubtedly did live and publish in a poetic milieu in which notions of decorum, balance and proportion, largely derived from the satires and epistles of Horace, Persius, and Varro, were desirable artistic goals. Therefore, ‘polished’ was, in the 1720s, far more likely to refer to the ‘beautiful’ than it would have in the 1790s. The noun ‘rudeness’ underwent a kind of reverse semantic evolution. In Pope's milieu it named verse which was distended and distorted by extremes of passion and unchecked imagination. But by the end of the century the noun named a redemptive energy; a force which could negate the stultifying learnedness and elegance of the ‘polished.’ Focusing now on the noun ‘beauty,’ a clear distinction can be made between the lexical company it keeps in the common repository of ECCO, and what is commonly found around ‘beautiful.’ A great deal of the language across the three time slices appears to articulate a certain version of feminine beauty: ‘fair,’ ‘charming,’ ‘enamour,’ ‘captivated,’ ‘Helen.’ The high ranking of the word ‘charms’ in all three periods is a strong indication that ‘beauty’ was something being considered under the desirous male gaze. Of course, ECCO is overwhelmingly a male authored domain. It is difficult, surveying all three period lists, to surmise anything other than that a male sense of feminine beauty is constructed using this lexis, and that the female is not agential in choosing this language. However, it must be borne in mind that we cannot inspect the counterfactual: there is no overwhelmingly female corpus against which to test the idea that this is a discretely male conception. It may indeed be that the version of feminine beauty which is apparent from these first lists was one produced and possessed as much by women as men. The fact remains however that this is untestable. One interesting feature here is how unchanging this picture is compared to the binding lists for ‘beautiful.’ ‘[B]eauty’ is suspended in a more stable lexical environment throughout the century of printed texts; a fact indicated by the fact that its list of common terms across all three decades is longer. But even though the exact vocabulary changes across the three historical lists, the same semantic sense is very strong: that this is an enamored male perspective onto female beauty. Surveying the search results across the whole century of ECCO data, it becomes clear that indeed ‘beauty’ and ‘charms’ are very commonly yoked together, particularly in publications whose objectives are the refinement of ladies' appearances and deportment. Examples are too numerous to mention, and this study does not aim to close read. Nonetheless one extract, from the preface to The Art of Beauty, or, a Companion for the Toilet. In Which the Charms of the Person are Considered and Explained by the intriguingly named Eminent English Physician at the Russian Court (1760), captures the morally prescriptive tenor of much of the writing that places the two words in textual proximity: Altho’ piety, modesty, virtue, good sense, and ingenuity, ought to be the chief objects of every woman’s attention; yet since the frailty of human nature inclines men, rather to listen to their senses than their judgment, it must be allowed an innocent at least, if not an [sic] necessary care in the fair sex to cultivate beauty. For this purpose was the following book written; namely, to enable the ladies to cultivate, and illustrate the charms which Nature hath given them. (Eminent English Physician at the Russian Court 1760, iii–iv) There is, in the above text and in the lists above, an intriguing intersection between language naming the development of flora, and that describing the female object. One may chart the concurrent flourishing and decline of the flower and the woman: ‘bloom,’ ‘blooming,’ ‘deformity, ‘fade,’ ‘fades.’ Some other interesting features occur further down the lists such as ‘garment’ and ‘cleanliness’ in the list for 1750s. The high ranking of ‘rebuke’ in all lists shows that beauty was certainly embedded in an ethics of sorts. Again, the intention of this article is not to close read, but one can imagine how beauty is something to be rebuked as a manifestation of pride and folly. We are in the era of Jonathan Swift's vituperations against female vanity and it appears likely that ‘beauty’ would form a locution with ‘rebuke’ in the sense that a woman's too great interest in beauty would (by some lights) require chastening. II. THE COMMON SEMANTIC NETWORKS FOR ‘BEAUTY’ AND ‘BEAUTIFUL’ These lists, produced by each single word, are inadequate if the aim is to reconstruct the structures of knowledge within which the noun and adjective were embedded in the eighteenth-century corpus. We may now, by computation, take each set of terms in which a word is embedded in each time period, produce binding lists for all of those words, and visualize the binding connections between many terms. Doing so will reveal the structure of the discursive environments which dominate the ECCO corpus in a range of historical periods. Figures 1 and 2 provide the first diagrammatical representation of this binding. They provide the semantic networks for ‘beauty’ and ‘beautiful’ respectively in the 1720–1730 time slice through ECCO. Figure 1 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beauty’ and related lexis in the 1720–1730 time slice of ECCO. Figure 1 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beauty’ and related lexis in the 1720–1730 time slice of ECCO. Figure 2 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beautiful’ and related lexis in the 1720–1730 time slice of ECCO. Figure 2 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beautiful’ and related lexis in the 1720–1730 time slice of ECCO. A clarifying note before interpreting: if words on these graphs are connected by a line, they appear on one or another's binding lists. This is known as first-order collocation or binding.10 The thickness of connecting lines indicates the strength of binding, quantified here as the dpf number between these terms. A thicker line indicates a higher dpf number.11 These semantic network diagrams straightaway confirm what was indicated by the initial binding lists, and expand the picture that they provided. Firstly, in the first network we find a clear visual representation of the fact that ‘beauty’ was indeed strongly yoked to ‘charms.’ Bound by a very strong line at the center of this diagram, these two words were focal points for the majority of this lexis in the 1720s tranche of ECCO. Dpf uses what is called a community detection algorithm to group words according to those which occur in proximity to one another most frequently in texts in this time slice. These communities are given colors, but the print edition of this article is black and white. Nonetheless we can observe the similarity of the lexis that connects to both ‘charms’ and ‘beauty.’ Remember that the thickness of the line joining these two words indicates that where ‘beauty’ occurred, so did ‘charms.’ Connected to both words is a list of the accoutrements and features of female deportment and the facets of male seduction. Again, the gaze appears highly gendered even if we cannot inspect a counterfactual, female corpus. Connecting to charms are ‘Chloe,’ ‘mien,’ ‘ravish’ ‘softness,’ ‘captivated,’ and ‘amorous.’ And the community with ‘beauty’ at its center, in the bottom half of this diagram, may be less overtly gendered, but still indicates sexual desire, male pursuit: ‘features,’ ‘witty,’ ‘sweetness,’ ‘admiring,’ ‘admiration,’ ‘attire,’ ‘charming,’ ‘ornament,’ ‘graceful,’ and ‘amiable.’ That suggestive adjective ‘deformed’ indicates the embeddedness in a discourse of female cosmetic and sartorial appearance that admitted sharp judgements of how appearances could be undesirable. Turning to visualizations of lexical co-association for ‘beauty’ in the decade 1750–1760, two features may be observed immediately. Firstly, the structured relations between words in each semantic network broadly retain the shapes that we observed in the earlier time slice. By this is meant that, broadly speaking, ‘beauty’ sits at the center of a plethora of qualifying lexis. It and ‘charms’ are still the anchoring words for a great deal of the vocabulary of female appearances (see Figure 3). Looking at this bicycle wheel shaped structure, and comparing it with the structures of “co-association for ‘beauty,’” further highlights differences in their usages and, we might begin tentatively to accept, the knowledge structures within which the noun and adjective were suspended (see Figure 4). Figure 3 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beauty’ and related lexis in the 1750–1760 time slice of ECCO Figure 3 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beauty’ and related lexis in the 1750–1760 time slice of ECCO Figure 4 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beautiful’ and related lexis in the 1750–1760 time slice of ECCO. Figure 4 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beautiful’ and related lexis in the 1750–1760 time slice of ECCO. This aforementioned, new type of difference between noun and adjective deserves some unpacking. These diagrams provide further evidence not only of the semantic separateness of these words, but of the ways in which the lexis around them is connected and structured. Put plainly, ‘beauty’ is situated at the heart of the wheel whereas ‘beautiful’ is one node in a less hierarchical, more interconnected network. In this middle decade as in the earlier one, Again, ‘beauty’ and ‘charms’ constitute two focal points for a wide range of words, and these words, on the whole, are not interconnected themselves. Contrast this with the network structure for ‘beautiful,’ and we may note how the lexis binding to ‘beautiful’ and ‘marble’ is itself interconnected, visualized in the crisscrossing lines between nodes. We will go into the implications of these different structures in the next section, and argue that these shapes signify differences in how knowledge was structured in these time periods of ECCO data. But for now, we will conclude on some points on the semantic, rather than the structural, contents of these networks. While it would be misleading to suggest that the domain of male and female romantic desire and pursuit, the world of cosmetic and sartorial charms, were unrelated to aesthetics in the anglophone eighteenth century, there is no significant overlap in these mid-century graphs and those for the 1720s, between an overtly artistic, artefactual lexis, and the vocabulary of the female toilet. One of the aims here is to facilitate comparison between the common and specialized uses of these words. And so we can with some confidence assert that the common semantic lives of ‘beauty’ in the ECCO corpus in the 1720s and 1750s do not resemble the semantic fields of philosophical aesthetic discourse. Irrespective of the fact that we know that aesthetics was of course a specialized discourse, it does nonetheless seem notable that there should be such a lacuna between its semantic suspension as illustrated above, and the world of artistic affect and aesthetic effect. The picture is quite different for ‘beautiful.’ Its common usage across the corpus in these two decades of texts shows to us a stark semantic distinction between the ‘beautiful’ materiality of sculpture and landscape, and that ‘beauty’ which is endearing, dazzling, captivating, and enamoring. What is fascinating is that beauty seems to operate, in these common early and mid-century semantic networks, as what Peter de Bolla has called a “containing concept.”12 That is, along with ‘charms’ with which it is lexically yoked, it seems to encompass and be comprised of, the paraphernalia of heterosexual courtship, as viewed from the male perspective. To be clear, beauty does not contain a typology; it is not comprised of a list of types of beauty. Rather, it functions taxonomically, arranging and structuring the apparatus and facets of male ideas of feminine attractiveness. By contrast, beautiful is surrounded if not by versions of itself as such, then at least by the materials from which ‘the beautiful’ could conceivably be fabricated. At least from the point of view of philosophers, a concept can be referred to by either or both substantives and adjectives, as well as substantives made from adjectives (‘the beautiful’). Therefore, the claim should not be that ‘beauty’ as such is not a concept of importance in eighteenth-century aesthetics. Here we run up against the seemingly intractable problem of slippage between word and concept. Just because a certain word is by a certain view not prominent, or is absent in a given discourse, this should not be taken as evidence for the concept's absence: its other lexical extensions may inform us of just how the concept is leaving its traces in language and shaping knowledge. The lexical surroundings of the adjective are things which humans make which are beautiful, and the materiality of such things. It has already been noted that its lexico-semantic surroundings articulate the tactile, physical arts of landscaping and sculpture. Painting seems subordinate to an aesthetics of the malleable; that which can be refashioned. Now, as we turn toward the picture provided for noun and adjective in the late eighteenth century, the focus will turn to the structural characteristics of each, and the remarkable changes that these undergo in the late eighteenth century. III. FROM CO-ASSOCIATION TO THE KNOWLEDGE STRUCTURES OF ‘BEAUTY’ AND ‘BEAUTIFUL’ To fully grasp the implications of the differences between ‘beauty’ and ‘beautiful’ for aesthetics, attention must be paid not only to the semantic contents of their lexical networks, but to the structures of relations within these. Indeed, this mode of attention, looking at shapes produced by lexical co-associations in addition to the semantic contents of the words that are co-associating, is one fascinating new aperture on word use and the structuration of knowledge that has been afforded by digital means of enquiry. In both the 1720s and 1750s networks we observed how ‘beauty,’ along with ‘charms,’ sat at the center of a bicycle wheel type shape of binding connections. These words were crucial convergence points for a welter of vocabulary, particularly, as we have just reconstructed by using a centrality measure, the noun ‘beauty.’13 Recognizing as much does not necessarily mean that ‘beauty’ or ‘charms’ were any more important than any other terms in these discursive environments of male desire, but that they were undeniably lexical and conceptual presences to which a considerable amount of other vocabulary gravitated. ‘[B]eauty’ and ‘charms’ were, self-evidently, touchstones without which the uses of the other words and their meanings would lose this particular type of consistency in this particular precinct of eighteenth-century printed knowledge. In the case of beauty, as has been articulated, (and one is aware of stretching a simile rather), ‘beauty’ and ‘charms’ operate almost like makeup bags or cabinets, containing the apparatus of female cosmetics and beauty. The words, and by extension the concepts, contain the tools and characteristics by which female beauty coalesces. In his pleasingly rebarbative book Graphs, Maps, Trees, Franco Moretti strikes a justly skeptical note at the start of the second chapter: There is a very simple question about literary maps: what exactly do they do? What do they do that cannot be done with words, that is: because if it can be done with words, then maps are superfluous. Take Bakhtin’s essay on the chronotope: it is the greatest study ever written on space and the narrative, and it doesn’t have a single map. Carlo Dionisotti’s Geografica e storia della letteratura italiana, the same. Raymond Williams’s The Country and the City, the same. Henri Lafon’s Espaces, romanesques du xiiie siècle … Do maps add anything to our knowledge of literature? (Moretti, 2007, 35)14 This is most definitely not a study of literature. Here, the objective is to provide a new or complementary account of the development and transmission of knowledge as viewed across a very large historical data set. To paraphrase, then, what exactly do the above kinds of network diagram do? Principally, they orient our attention around structural relations between numerous words. Thinking this way about the above diagrams, what has been made visible by them, is that the words ‘beauty’ and ‘charms’ behave as central points or hubs for related lexis. The words are the unifying forces in a lexical community: other terms qualify, add richness or complexity to, the central term. As already noted, the lines (called edges in network theory vocabulary) which flow into and out of ‘beauty’ look something like the spokes of a bicycle wheel. This wheel like structure (hereafter called a degree community) is very common in the network diagrams which are produced from ECCO data, but it is only one shape among several. These patterns are the grammar of impersonal knowledge. They are signals, observable only when one can scan across the whole of ECCO as we can here, about how one language use, and by extension knowledge, was structured. Digital enquiry is allowing us a new view onto a more structural understanding of the vast diachronic shapes and movements of how many people knew ‘beauty’ and the ‘beautiful.’ To appreciate this idea of a knowledge structure further, we might imagine the counterfactual—that instead of ‘beauty,’ or indeed any word, sitting centrally amidst a host of lexically-bound word nodes, there was mutual or reciprocal binding between all nodes in a given part of a graph. Instead of one word being a convergence point for many other words’ binding, all word nodes connect to all others.15 In fact we do not need to imagine, but may observe what this looks like (see Figure 5). This example comes from the already cited domain of logarithmic mathematics, this time across the later data set of 1770–1780 in ECCO. Figure 5 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beauty’ and related lexis in the 1720–1730 time slice of ECCO. Figure 5 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beauty’ and related lexis in the 1720–1730 time slice of ECCO. Many, many authors in this historical period of the eighteenth century used this set of words and their meanings, in a pattern of mutual interconnectedness. All words are reliant on all other words: this is an epistemological structure which operates as a set rather than with repeated reference to one term in particular. This will be called a “relational community.” As already noted with reference to the adjective ‘beautiful,’ this structure suggests a more diffuse or equal distribution of use and emphasis across several terms, rather than a reliance on one.16 We began to note the following with relation to centrality in the previous section: if ‘beauty’ is central to the degree community, along with ‘charms,’ then the words around ‘beautiful’ show a far greater tendency toward interrelatedness. Figures 6 and 7 are cropped—the first from 1750 to 1760 for ‘beautiful’ and the second from 1720 to 1730. Both contain the crisscrossing that is so pronounced in the above case of mathematical terminology. This is clear evidence that ‘beautiful’ in the early and middle century of ECCO data is not only semantically distinct from ‘beauty,’ but that the ways in which its co associating lexis relates, is structurally distinct, too. A set of terms which tend to relate to several of the words in that set, as we see in these two cases above, is a different type of language use to that in which numerous terms relate only to one centralizing word node. ‘Beautiful’ was a word surrounded by relational discursive communities of artistic production; the grit and touch and work of sculpture and landscaping. These communities came with their own internal logic and terms of reference, whereas the words surrounding ‘beauty’ and ‘charms’ tended only to need these nouns for certain types of (amorous, cosmetic) sense to be made. We can well imagine how ‘granite,’ ‘porphyry,’ ‘marble,’ and ‘variegated’ combine in a discourse of ‘the beautiful,’ without that adjective containing these as such, as we saw in the case of ‘beauty’'s containment of the accoutrements of female appearances. This is evidence that knowledge structures constituted of ‘beauty’ and ‘beautiful’ were different in semantic and structural terms. The semantic distinctiveness of both, in addition to and working in agential concert with, the very clear structural differences laid out above, point to two quite different ways of knowing. Figure 6 Open in new tabDownload slide Cropped image of Figure 4 (‘beautiful’ in 1750s), showing strong and mutual lexical co-associations. Figure 6 Open in new tabDownload slide Cropped image of Figure 4 (‘beautiful’ in 1750s), showing strong and mutual lexical co-associations. Figure 7 Open in new tabDownload slide Cropped image of Figure 2 (‘beautiful’ in 1720s), showing strong and mutual lexical co-associations. Figure 7 Open in new tabDownload slide Cropped image of Figure 2 (‘beautiful’ in 1720s), showing strong and mutual lexical co-associations. What is most remarkable, however, is the change that can be observed as we reconstruct the semantic networks for noun and adjective in the final decade of ECCO. The vast upswing in printed materials in the time (and consequently in the corpus), effects great structural changes (see Figures 8 and 9). This is nothing less than a virtual reversal of the structural characteristics which were observable in the semantic networks for the early and middle periods of ECCO data. In stark contrast to those earlier semantic networks, the words in this environment for ‘beauty’ are now interrelated to a significant degree, with the community on the right, including ‘charms,’ ‘bewitching,’ ‘captivating’ and so forth, in particular now forming a set of terms bound by several interrelated links. Far from seeming a superordinate focal point, ‘charms’ has emerged in common use at the end of the century of print in Britain, as one highly connected node in a relational community of terms relating to graces and enchantment (see Figure 10). Figure 8 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beauty’ and related lexis in the 1790–1800 time slice of ECCO. Figure 8 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beauty’ and related lexis in the 1790–1800 time slice of ECCO. Figure 9 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beautiful’ and related lexis in the 1790–1800 time slice of ECCO. Figure 9 Open in new tabDownload slide Semantic network showing lexical co-associations for the word ‘beautiful’ and related lexis in the 1790–1800 time slice of ECCO. Figure 10 Open in new tabDownload slide Cropped image of Figure 8, showing strong and mutual lexical co-associations. Figure 10 Open in new tabDownload slide Cropped image of Figure 8, showing strong and mutual lexical co-associations. As already acknowledged, the words surrounding the search term, and relations between these, contribute to that word’s meaning. Therefore, in this final decade of ECCO printed material, this precinct of the lexis around ‘beauty’ is self-evidently more internally related, more diffuse and shared, than in previous time slices. The vast polity of authors in this period of print’s great anglophone proliferation, have produced texts which discuss female attractiveness more as an interrelated set of words and concepts, rather than primarily with recourse to, or binding to, the word ‘beauty.’ This is a change that we have observed through time: ‘beauty’ is no longer an indispensable touchstone, for this part of the discourse at least. Similarly, in the entangled communities in the bottom of this diagram, we notice that these words are so interrelated as to cause the community detection algorithm some difficulty. ‘Deformed,’ ‘deformity,’ ‘wrinkles,’ ‘complexity,’ ‘features,’ ‘bloom,’ ‘fade,’ and ‘fades’ now relate mutually in a way that they did not in the earlier semantic networks. Therefore, again, neither ‘beauty’ nor ‘charms’ have the same centrality that we observed earlier in the ECCO data: new interrelations have emerged in this last decade, new avenues for knowledge have occurred which do not rely to the same extent on the use of a single word. And if the picture of ‘beauty’ is more democratic, more relational, the network for ‘beautiful’ has moved in the opposite direction. The adjective is now suspended in a semantic network in which it is clearly the central node. Where in the past it was accompanied in this centralizing role by the world ‘marble’ in the earlier networks, the adjective now sits alone, less a part of a network of interrelated words than a qualifier for numerous unconnected words. If we compare these cropped, close-up images of the terrain around ‘beautiful’ in 1750 and then 1790 we get an immediate impression of the new centrality of the adjective; its late-century function as a word which does not so much sit among many related nodes, but which qualifies many words which may or may not interrelate (see Figures 11 and 12). Figure 11 Open in new tabDownload slide Cropped image of Figure 6, highlighting interrelated co-associations between words in this semantic field. Figure 11 Open in new tabDownload slide Cropped image of Figure 6, highlighting interrelated co-associations between words in this semantic field. Figure 12 Open in new tabDownload slide Cropped image of Figure 9, highlighting lack of interrelations between words in the 1790–1800 time slice for the word ‘beautiful’. Figure 12 Open in new tabDownload slide Cropped image of Figure 9, highlighting lack of interrelations between words in the 1790–1800 time slice for the word ‘beautiful’. In addition to this new centrality for ‘beautiful,’ another new feature of this later period of data is that the semantic network within which it sits so centrally, is sparser than in the early and middle decade of ECCO. This seems counterintuitive. There are immeasurably more texts in this decade of ECCO than there are in the earlier two time periods. And so this new sparsity demonstrates that numbers of texts in corpora and frequencies of words per se, do not necessarily correlate to denser word co-associations. A word can rise massively in use while the semantic domains within which it is used can become more limited. This is the case for ‘beautiful,’ where the numbers of words that it qualifies have certainly contracted despite a surge in publications using the word. In other words, as the number of texts has increased, ‘beautiful’ has become less diverse in its connections, and connected in a different way to how it functioned before. Again, we may think about this relative to its uses in aesthetic treatises—do we note a similar contraction in the diversity of its semantic fields? The semantic network for the final decade is showing us that the word’s function has become more markedly qualificatory, and less connective. In 1790–1800 it describes features of artistic domains without connecting those domains. Various domains of artifactual production are still adjusted by ‘beautiful’ but there is less interchange and connectedness between, say, the worlds of landscaping and painting. The communities in this later network are neater and more delineated. It is now easier to observe the separateness of the architectural community in blue in the top right of the network, than it was in previous semantic networks. Once again, this is an avenue for further enquiry: does this view onto the common uses of the adjective align with how the word is used in essays on taste, landscaping and painting? Another interesting factor in this view onto the last decade of ECCO, is that the semantic contents of these later graphs are virtually identical to those in the earlier parts of the century. As was the case with ‘beauty,’ what have changed are the structured relations between bound words in each semantic network, not the words themselves. With the addition of some of the language of the decorative arts and a general decline in the semantic diversity overall, the picture for ‘beautiful’ is still recognizably aesthetic. Similarly, ‘beauty’ is still embedded in the domains of attraction and desire, even if the communities of terms surrounding this word are marked by a new relationality than they were in the earlier century. IV. CONCLUSION: COMPARING THE COMMON TO THE SPECIALIZED Let us return to the question of which claims seem feasible and which unfeasible in light of the above results. Decidedly, the above work and findings do not constitute the basis for a revision of narratives of aesthetics, nor even a remediation of the semantic work that ‘beauty’ and ‘beautiful’ were doing in the aesthetic corner of British philosophy in the eighteenth century. These are not reasonable aims because the object of enquiry in this article has been the aggregated whole of print as held by the ECCO corpus: not the specialized and relatively tiny set of texts dealing with what might be called “aesthetic beauty.” Nevertheless, we can and indeed should reflect upon the convergences and divergences between what we know about aesthetics from centuries of close reading that relatively small philosophical constituency, and what we have begun to uncover using computation above. Today, scholars of aesthetics making enquiries about beauty will encounter, in material returned by the most cursory internet searches, to more protracted scholarly philosophical enquiries, assertions along the lines of “Frances Hutcheson’s conception of beauty was that it was ‘uniformity amidst variety.’” They will encounter claims that Anthony, Earl of Shaftesbury was the progenitor of the idea of disinterestedness in aesthetic experience of beauty, and that this idea would undergird assertions of taste from one end of the eighteenth century to the other. By such acts of transmission, undoubtedly, were some keenly appreciated semantic contents of the word ‘beauty’ given definition. But it seems perverse not to acknowledge, in addition to the brilliance of these transmissions, the wider story of ‘beauty’ as a way of knowing for most people writing and reading in this century. There is significant evidence above to suggest that beauty was linked to the world of heterosexual desire in the common stock of printed knowledge. In its most common uses in print, ‘beauty’ was primarily a concept not of wrought artifacts or striking natural phenomena, but a noun of the drawing room and toilet. That this is true, is nothing more or less than a logical extrapolation from fairly simple lexical co-association enquiries into the historical textual data set; it is not an ideologically driven assertion, nor a challenge to how ‘beauty’ is used in aesthetics. What is perhaps more suggestive and intriguing is that in its common uses across ECCO as a whole and indeed across the century, the adjective ‘beautiful’ coincides with philosophical aesthetics where ‘beauty’ does not. They are clearly semantically distinct. Reading the following, from Agnes Heller's fine discussion of taste in Adam Smith and David Hume, one gain a sense of how much this type of historical account feels insensitive to facts and implications of word type and conceptual discreteness: One can still detect a dominant concept of the beautiful in Hume since the move of negation— that is, “This is not beautiful, but …” —occupies a central space in his reflections on the distinction between taste and delicate taste. However, in Adam Smith’s philosophical ruminations, no such concept appears to be preserved or discussed. Smith seems adamant that there is no measurement by which to compare beauty in different things. His primary thesis holds that all kinds of beautiful things are beautiful in and of themselves, by their own intrinsic measure. In other words, the means by which one measures beauty in something like furniture naturally differs from the means by which one measures beauty in nature. (Heller 2012, 38–39) In light of the above research, we can now compare this kind of argument to what we know about the traffic of knowledge more widely using these words more common traffic. Heller refers to Smith’s thinking on “beauty in different things” but for most readers, as has been shown, ‘beauty’ itself is a different thing, to ‘the beautiful.’ Given the clear separateness of these two words in terms of semantic embeddedness and structural relations to other associated words, it seems secure that “beauty in something like furniture” would strike the common reader's ear as an unusual observation. The evidence above indicates that it would feel odd to most writers and readers to see a piece of furniture described as ‘beautiful.’ The wider, most common set of cultural practices and assumptions freighted with these words are only now coming to light because of digital enquiry, and going forward it would show the right kind of historical fidelity to acknowledge the special and particular ways in which aesthetics was carving out a new and strange semantic space for the word ‘beauty.’ That strange space was cleared in a now familiar sematic journey in the aesthetic corner of philosophy. In this, the word ‘aesthetic’ did not so much change its meaning, as accrue one particular significance which is still with us today. In 1735, when Alexander Gottlieb Baumgarten ([1735] 1954) used the word, it referred mainly to the empirical study of sense perception and the imagination. Gradually over the years, passing through the theoretical hands of Shaftesbury and Francis Hutcheson, from Gottfried Wilhelm Leibniz and Christian Wolff to David Hume and Immanuel Kant, the word (so the usual narrative goes) ended the European century also referring to the study of beauty, whether this encompassed natural or artifactual beauty, or both.17 Of course, the concept of beauty existed prior to the semantic change wrought in and by the word ‘aesthetic.’18 Nonetheless in aesthetic discourse in the eighteenth century, the two words' semantic evolution were intertwined. Once again, in light of the above, we may begin to tell this story against the more shared space of virtually all printed material. We have seen how the word ‘beautiful’ is surrounded by the language of sculpture, architecture and landscaping. Language which seems specific to painting, dance, music and poetry, does not feature in these lists of bound terms. The claim here is not that painting, dance, music and poetry were unimportant to aesthetics. Rather, it is that these do not appear, on this not unsubstantial first evidence, to inflect the common discourse of the ‘beautiful’ to anywhere near the same extent as sculpture, architecture and landscaping. If the ‘beautiful’ is a concern in aesthetics then, again self-evidently, it is the case that the art of marble and granite, that which is stately and variegated, that is described most often in the thousands of uses of that word across the e-historical corpus. Investigating knowledge as it is held in the impersonal, aggregative repository of the historical corpus, differences in word type have here been used as apertures onto different ways of knowing in the most common sense.29 What has been made visible are the knowledge structures that obtained across the vast, impersonal whole of the ECCO corpus and invisible to close readers, no matter how many these number or how assiduously they read. Associations between many, many words, the semantic communities which are generated, by these, the patterns and forces within them, and the structures of knowledge which come about as these are used by many authors, are indissolubly linked. The processes of historical reconstruction in this study would have been be impossible without the affordances of digital technology.20 REFERENCES Alonso , Araceli , Chrystel Millon, and Geoffrey Williams. 2011 . “Collocational Networks and Their Application to an E-Advanced Learners Dictionary of Verbs in Science (DicSci).” In Electronic Lexicography in the 21st century: New Applications for New Users: Proceedings of eLex, Bled, 10–12 November 2011 , edited by I. Kosem and K. Kosem, 12 – 22 . www.trojina.si/elex2011. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Baker , Paul . 2016 . “The Shapes of Collocation.” International Journal of Corpus Linguistics 21 : 139 – 64 . Google Scholar Crossref Search ADS WorldCat Baumgarten , Alexander Gottlieb . (1735) 1954 . Reflections on Poetry: Alexander Gottlieb Baumgarten’s Meditationes philosophicaede nonnullis ad poema pertinentibus . Translated by Karl Aschenbrenner and William B. Holther. University of California Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Brezina , Vaclav , Tony McEnery, and Stephen Wattam. 2015 . “Collocations in Context: A New Perspective on Collocation Networks.” International Journal of Corpus Linguistics 20 : 139 – 73 . Google Scholar Crossref Search ADS WorldCat Church , Kenneth Ward , and Patrick Hanks. 1990 . “Word Association Norms, Mutual Information, and Lexicography.” Computational Linguistics 16 : 22 – 9 . Google Scholar OpenURL Placeholder Text WorldCat De Bolla , Peter . 2013 . The Architecture of Concepts: The Historical Formation of Human Rights . Fordham University Press . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC De Bolla , Peter , and Andrew Ashfield, eds. 1996 . The Sublime: A Reader in British Eighteenth-Century Aesthetic Theory . Cambridge University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Durrant , Philip , and Alice Doherty. 2010 . “Are High Frequency Collocations Psychologically Real? Investigating the Thesis of Collocational Priming.” Corpus Linguistics and Linguistic Theory 6 : 125 – 55 . Eminent English Physician at the Russian Court . 1760 . The Art of Beauty, or, a Companion for the Toilet. In Which the Charms of the Person are Considered and Explained . London : J. Williams . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Evert , Stefan . 2005 . “The Statistics of Word Cooccurrences Word Pairs and Collocations.” PhD diss., University of Stuttgart . https://elib.uni-stuttgart.de/bitstream/11682/2573/1/Evert2005phd.pdf. Glückler , Johannes , Emmanuel Lazega, and Ingmar Hammer, eds. 2017 . Knowledge and Networks . Springer Open Access Online: https://www.springer.com/gp/book/9783319450223. Gregor , Mary J . 1983 . “Baumgarten’s ‘Aesthetica.’” The Review of Metaphysics 37 : 357 – 85 . Google Scholar OpenURL Placeholder Text WorldCat Heller , Agnes . 2012 . The Concept of the Beautiful . Plymouth : Lexington Books . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Manning , Christopher D. , and Hinrich Schutze. 2005 . Foundations of Natural Language Processing . Cambridge, MA: MIT Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Moretti , Franco . 2007 . Graphs, Maps, Trees: Abstract Models for a Literary History . London : Verso . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Moretti , Franco , and Oleg Sobchuk. 2019 . “Hidden in Plain Sight.” New Left Review July/August 2019 . https://newleftreview.org/issues/II118/articles/franco-moretti-oleg-sobchuk-hidden-in-plain-sight. Phillips , Martin . 1985 . Aspects of Text Structure: An Investigation of the Lexical Organisation of Text . Amsterdam : North-Holland . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC ———. 1989 . Lexical Structure of Text . Birmingham University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC ———. 2001 . Words and Phrases . London : Blackwell . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Recchia , Gabriel , and Paul Nulty. 2018 . “Improving a Fundamental Measure of Lexical Association.” Proceedings of the 39th Annual Meeting of the Cognitive Science Society. https://pdfs.semanticscholar.org/c5e8/95600465165a6aeb7096c65fd515d7d6e848.pdf. Regan , John . 2018 . Poetry and the Idea of Progress 1760–1790 . London : Anthem Press . Stubbs , Michael . 1995 . “Collocations and Semantic Profiles: On the Cause of the Trouble with Quantitative Studies.” Functions of Language 2 : 23 – 55 . Google Scholar Crossref Search ADS WorldCat Footnotes 1 A related issue concerns the use of cognate words in French and German during the same period. The author intends the current work to be a conduit to further research. Going forward the corpus-investigation technology and interface employed here will, in time, be used to explore uses and associations for the relevant words in the German-language Deutsches Textarchiv (DTA) and the French Frantext corpus of the eighteenth to twentieth century (among others). 2 Throughout, the adjective ‘common’ is taken to mean the privileged few who were publishing and editing in Britain in the eighteenth century. 3 This measure of lexical co-association was developed by the author and colleagues in a research project called The Cambridge Concept Lab at the University of Cambridge between 2012 and 2018. The lab's members were John Regan, Peter de Bolla, Ewan Jones, Gabriel Recchia, and Paul Nulty. 4 As in many measures of this kind, for dpf, the “observed” number of co-associations between words in the corpus is compared with an “expected” baseline. This baseline is in fact an artificially contrived comparator in which no word is more or less likely to co-associate with any other word. In other words, the actual binding between words in natural language is compared with an artificial “expected” number in which all terms in the corpus are randomised, to calculate strength of binding between words. The headline benefit of the method used here in particular, is that it nullifies the impact of a word's frequency in the corpus as a factor determining its dpf score. This means that very infrequent words do not immediately get very high dpf scores, and very frequent words do not get low dpf scores. 5 The list is curtailed—the full list would run for tens of pages. One of the perennial problems of the user of data in enquiries such as these is to decide how much to use, where to threshold. There is no settled view on how many words are available to the individual thinker as they undertake the various forms of linguistic communication. But more pertinently, this study does not concern individual thinkers but in fact wishes precisely to escape their cognition, to recover, historically, what a vast polity of writers wrote. There is certainly no theory of how many words should make up a list such as this, therefore, and I have used my discretion (and sense of page size) to make the cut off at thirty words. 6 In the overwhelming majority of cases, the word most likely to occur with ‘logarithm,’ is ‘logarithm.’ This is a phenomenon known to corpus linguists as “burstiness,” and this is particularly pronounced here. This is to say, that ‘logarithm’ recurs over and over again near itself in mathematical writing in this time slice of ECCO data. 7 Many features of these communities contribute to how meaning is generated. These include the semantic contents of the words which constitute them, strengths of binding between those words, most and least frequent words in the community, and several other factors. An important innovation in lexical collocation or several words is Evert (2005); another is Chapter 5 of Manning and Schutze (2005), which has a clear explanation of various measures, and contains a very clear explanation of why low frequency words cause a problem for PMI in Section 5.4. Finally, a brief but nonetheless useful guide is Recchia and Nulty (2018). 8 All such decisions (about which time periods to choose, how long to make lists of bound terms and so forth) are based on the discretion of the researcher. There are benefits and drawbacks to every choice made. 9 For a fuller investigation of the different directions of travel for ‘polished’ and ‘rudeness’ in eighteenth-century aesthetics see Regan (2018). 10 What we are being shown in these graphs is the likelihood of word use among a great many writers in a given time period. Each word node is like a junction on a road map. As in most road maps, we see a delimited terrain and a limited number of moves that one may make given where one is situated. The many thousands of eighteenth-century writers are the travellers in this environment, and the graphs reflect the many thousands of decisions that they made about word choices in their writing. If there is a line between two words, a great many writers chose to use one word in a sentence with another. 11 These interpretations of network diagrams, collocations and network theory in relation to knowledge are informed by the following: Stubbs (1995), Philips (1985, 1989, 2001), Brezina et al. (2015), Church and Hanks (1990), Durrant and Doherty (2010). 12 This theory of containing concepts, along with a more fully realised typology of concepts generally, is developed in de Bolla (2013). 13 Arguments about the relation of network motifs to knowledge structures, while not profuse in the literature, have received some attention, as in Alonso et al. (2011). 14 Equally relevant to this part of my discussion is Moretti and Sobchuk (2019). 15 My understanding of knowledge and networks, with particular emphasis on structural and spatial elements, has been shaped by Glückler et al. eds (2017). 16 Relations between word binding and the structures of knowledge are discussed suggestively in Baker (2016). 17 The most compelling and lucid account of this historical change is provided in the introduction to de Bolla and Ashfield, eds (1996). Gregor (1983) also provides a fine historical account of Kant's critique of rationalist aesthetics and the provenance of aesthetics in logic. 18 Italics will denote references to concepts as opposed to words as they appear in the ECCO corpus. 19 Of course, the argument could be presented with the agency inverted. Reversing the dynamic, one could begin with the position: the contours of what was known about art and the sublime, truth and knowledge, were in part shaped by the different grammar employed in signifying the human experience of art and the sublime. Different word types produced distinctive forms of language use and, by extension, particular structures of conceptualisation. In regarding a corpus as a repository of knowledge, one cannot consider the corpus as a kind of brain to which mental representations (concepts) are made. Starting with the idea of concepts as shared, common entities, language use does not merely give the historian of ideas signals of conceptuality; it constitutes and shapes that conceptuality. The author does not necessarily eschew this position, but in plain terms it is not the concern of the current enquiry. 20 The author would like to acknowledge the work of his colleagues in The Cambridge Concept Lab, with whom the measure of lexical co-association, and the visualisation tool used in this article, were designed and built. Thank you to Peter de Bolla, Ewan Jones, Gabriel Recchia and Paul Nulty. Thanks also to all of the researchers at Cambridge’s Centre for Research in the Arts, Social Sciences and Humanities (CRASSH), faculties of English and History, and Clare Hall in particular who contributed their time and critical attentions to these digital tools so generously over the years. © The Author(s) 2021. Published by Oxford University Press on behalf of the American Society for Aesthetics. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - ‘Beauty’ and the ‘Beautiful’: a Computational Analysis of the Company They Kept Across the Eighteenth-century Corpus
JF - The Journal of Aesthetics and Art Criticism
DO - 10.1093/jaac/kpaa008
DA - 2021-02-19
UR - https://www.deepdyve.com/lp/oxford-university-press/beauty-and-the-beautiful-a-computational-analysis-of-the-company-they-AfIu2NtakM
SP - 88
EP - 107
VL - 79
IS - 1
DP - DeepDyve
ER -