‘Three-quarters, kit-cats and half-lengths’: British portrait painters and their canvas sizes, 1625-1850

2. Four historic sizes

Between the 1620s and the 1690s, the basic range of British standard sizes was established in London: the half length, the three-quarters, the kit-cat and the whole length.

2.1 The half-length

In 1632 Van Dyck was paid by Charles I for three portraits ‘at half length’ of the royal family, as well as for other portraits ‘at length’. [56] As a term, ‘half length’ was used in a listing of restored pictures in the Clarendon collection, c.1683-5. [57] The half-length as a description was an international term and implied a portrait to below the waist and generally to the knees. Indeed such portraits are sometimes so described in early sources in Britain: ‘to ye knees’ or a ‘knee cloth’, though in most inventories and descriptions the term ‘half-length’ was preferred.[58] Such descriptions remained in common currency, used for example by Sir Joshua Reynolds, already quoted, who described a portrait to a client as ‘As far as the Knees’. In 1792 the poet William Cowper described his seated portrait by Lemuel Abbott (fig.5), ‘It is half-length, as it is technically but absurdly call’d: that is to say it gives all but the foot and ancle’. [59] Since the mid-19th century, the half-length has generally been described as a three-quarter length.

Half-length format. Lemuel Abbott’s William Cowper, 1792, 50 x 40 ins (127 x 101.6 cm) (National Portrait Gallery). Cowper wrote of this portrait, ‘It is half-length, as it is technically but absurdly call’d: that is to say, it gives all but the foot and ankle’.

Fig.5 Half-length format. Lemuel Abbott’s William Cowper, 1792, 50 x 40 ins (127 x 101.6 cm) (National Portrait Gallery). Cowper wrote of this portrait, ‘It is half-length, as it is technically but absurdly call’d: that is to say, it gives all but the foot and ankle’.

In England the ‘half-length’ was well on the way to being standardised at 50 x 40 ins by the 1650s. While Van Dyck’s half-lengths of the 1630s were usually about 41 x 33 ins, but sometimes a few inches larger, and William Dobson’s of the 1640s about 44 x 36 ins, with a few as large as 49 x 40 ins, Peter Lely moved away from Van Dyck’s format to one approaching 50 x 40 ins from about 1650. [60] An analysis of half-length portraits in the collection of the National Portrait Gallery shows a majority at or just under 50 x 40 ins in the 1650s and a great majority by the 1680s. [61] The use of canvas of this size had become so much the norm during the course of the 17th century and early 18th century that when Jonathan Richardson chose otherwise in the 1730s it was a matter for particular comment by George Vertue: ‘lately he makes his Cloths of his half length. 3 Inches longer than others & 1 inch ½ wider’. [62]

In February 1677 Charles Beale purchased 20 ¼ ells of good canvas to make 18 half-length cloths for painting by his wife, Mary Beale, that is 50.6 ins per cloth. [63] Although this would seem tight for a 50 x 40 ins picture, in buying the canvas Beale may have received an extra inch for each yard or ell (see section 1.2, especially note 18) and in stretching it he will have obtained some give from the material.

Turning to the 1730s, an unlined 50 x 40 ins portrait of Alice Caulfield, daughter of William 1st Viscount Charlemont, attributed in the saleroom to Joseph Highmore but more likely to be by Bartholomew Dandridge (Christie’s 30 November 2001 lot 9), with selvedges on the long sides, had 2 ins turnovers on each side, giving a roll width after stretching of about 44 ins. [64] A slightly narrower roll width of 41 ½ ins can be determined for Lemuel Abbott’s William Cowper, 1792 (fig.5) from the presence of the numerals, 115, in the relevant compartment of the ‘frame mark’ on the reverse (as explained in section 1.2).

Eventually two extra half-length sizes came into the repertory of standard sizes: the large half-length, size 56 x 44 ins, called a Bishop’s half-length, which is described more fully in section 3.1 below, and the small half-length, size about 44 x 34 ins. This smaller size, or something like it, seems to have used with variations for much of the 17th century and into the early 18th century. It may be what Charles Beale called an ‘ell cloth’ (his size seems to have been about 43 x 34 ½ ins) and John Smibert an ‘O.S.’ (meaning unknown) or a little half-length. [65] It was not until the early 19th century that the small half-length became formalised (see fig.3) and even then, though used by three leading portrait painters, James Northcote and Samuel Drummond in London and Thomas Sully in America, it did not find general favour. [66] As a size it was a return to something close to the format that Van Dyck and his immediate predecessors had favoured.

Three-quarters format. Nathaniel Hone’s Kitty Fisher, 1765, 29 ½ x 24 ½ ins (74.9 x 62.2 cm) (National Portrait Gallery).

Fig.6 Three-quarters format. Nathaniel Hone’s Kitty Fisher, 1765, 29 ½ x 24 ½ ins (74.9 x 62.2 cm) (National Portrait Gallery).

2.2 The three-quarters

Canvases for head-and-shoulders portraits, sometimes called bust-lengths, seem from observation to have been standardised among some London portrait painters during the later 1620s and 1630s at or about 30 x 25 ins (fig.6). In the case of Cornelius Johnson, he seems to have abandoned his panel format of about 26 x 19 ½ ins in the mid-1620s when he began using canvas regularly. He then chose to adopt sizes of about 29 x 24 ins or 30 x 25 ins both for his canvases and for his rather fewer panels, perhaps in response to an increase in scale in the work of his rivals. [67] The term ‘three-quarters’ as a description of size first appears in inventories in the 1640s and 1650s, as for example in the list of pictures in Van Dyck’s estate compiled in about 1649 (‘The Queens pictures 2; 3 quarters’ and ‘2 Italian Pictures 3 quarters’) or in the note Sir Ralph Bankes compiled in the late 1650s of his pictures (‘Dutch piece on a 3 qr Cloth’ and ‘A Church Piece of Prospective on A 3qr long Cloth’). [68] This is not to suggest that these pictures were necessarily precisely 30 x 25 ins but rather that ‘three-quarters’ was by the mid-century understood as a term for a size of picture. [69]

A number of explanations have been advanced for the use of this term: that such a canvas is three-quarters the size of a kit-cat,[70] that it is three-quarters of a half-length canvas,[71] that it is a portrait three-quarters the size of life,[72] that it is three-quarters of a square yard,[73] and that it is three-quarters of a yard in length. [74] Technically it is correct to say that a 30 x 25 ins canvas in surface area is three-quarters the size of a kit-cat - the actual figure is 74.4% - but since the kit-cat size did not come into use until many years after the three-quarters the explanation fails. Technically, too, it would be possible to make two 30 x 25 ins canvases from a 50 x 40 ins half-length but this would be very wasteful and would not have been acceptable in practice. The third theory, that a three-quarters is three-quarters the size of life, is clearly mistaken, as is the fourth that it is three-quarters of a square yard: in area it is only 57.9% of a square yard. It is the final explanation that is correct, as a note by the late 18th-century portrait painter, Ozias Humphry, confirms. He records that a three-quarters is a canvas of three-quarters of a yard in length, that is 27ins. [75]

This figure of 27 ins is the amount of material which would generally be allowed to produce a 25 ins wide canvas, taking into account the canvas turnover (the extra inch or so needed on each side for stretching the canvas and fixing it to the wooden strainer or stretcher). Information can be gained from the few unlined canvases that retain their selvedges (the woven edge of the cloth). The example of a portrait of Henry Oxenden, dated 1640 and once attributed to Cornelius Johnson, is instructive. Not only is this portrait unlined, it retains its selvedges at top and bottom. The painting has a stretcher of size 30 5/8 x 25 1/4 ins. Including the turnovers, the average size of the canvas is about 33¼ x 27 ¼ ins, that is a roll width of about 33 ins from which a length of about 27 ins was cut and then stretched. [76] The more common alternative, as we shall see, was to use a roll of about 27 ins wide from which a length of about 32 ins would be cut (fig.7).

Canvas for a three-quarters. A portrait size 30 x 25 ins can be cut from a narrow roll, about 27 ins across (left, with selvedges, if surviving, at top and bottom) or from a wider roll, width about 32 ins (right, with selvedges, if surviving, at left and right). The narrow area between the portrait and the canvas edge forms the turnover. Mrs Cibber and Alexander van Aken, both by Thomas Hudson, are used here as illustrations without evidence as to how they were actually cut (National Portrait Gallery). Not to scale.

Canvas for a three-quarters. A portrait size 30 x 25 ins can be cut from a narrow roll, about 27 ins across (left, with selvedges, if surviving, at top and bottom) or from a wider roll, width about 32 ins (right, with selvedges, if surviving, at left and right). The narrow area between the portrait and the canvas edge forms the turnover. Mrs Cibber and Alexander van Aken, both by Thomas Hudson, are used here as illustrations without evidence as to how they were actually cut (National Portrait Gallery). Not to scale.

Fig.7 Canvas for a three-quarters. A portrait size 30 x 25 ins can be cut from a narrow roll, about 27 ins across (right, with selvedges, if surviving, at left and right) or from a wider roll, width about 32 ins (left, with selvedges, if surviving, at top and bottom). The narrow area between the portrait and the canvas edge forms the turnover. Mrs Cibber and Alexander van Aken, both by Thomas Hudson, are used here purely for illustrative purposes, without evidence as to how they were actually cut (National Portrait Gallery). Not to scale.

Taking a sample of twelve unlined canvases of three-quarters size, we find eight with selvedges on both left and right sides, and so from a narrow roll of about 27 ins wide, and four with selvedges at top and bottom, that is from a somewhat wider roll of about 32 ins wide. [77] The figures for roll widths given here are after stretching. In chronological order those taken from narrow rolls, and so with the short side cut out across the roll, are works by Michael Dahl (c.1700, 25 3/8 ins roll width), Jean Baptiste van Loo (c.1737-42, 28 ¼ ins), James Northcote (1782, 27 ¾ ins), John Hoppner (early 1800s, 26 ins), Robert Home (1810s, 27 ¾ ins), Henry Raeburn (c.1818, 28 ins), Thomas Stewardson (1824) and John Simpson (exh.1833, 27 ¾ ins). [78] Those from wider rolls, and so with the long side cut out across the roll, are attributed to Cornelius Johnson (1640, 33 ¼ ins, see above), Joshua Reynolds (c.1745, 31 3/8 ins), George Romney (c.1787, 32 1/16 ins) and John Hazlitt (c.1818). [79]

The interest of these results lies not in the precise measurements, which may only be reliable to within an inch, but in the fact that over the centuries canvases were taken both from across and along the canvas roll according to its width. It is possible to supplement this data from two other sources: Charles Beale’s diary notebooks of 1677 and 1681 and the duty ‘frame mark’ found on commercially prepared canvases from 1784.

In 1681 Charles Beale produced two batches of three-quarter size canvas from ‘excellent Royston sacking’, the first in February of twenty-one canvases from a 16 yard roll, implying 27.43 ins per canvas, assuming no wastage, and the second in August of twenty-four canvases from an 18 yard roll, that is 27 ins per canvas. [80] The short sides of his canvases were evidently taken from the length of the roll, which may have been about 32 ins wide.

The significance of the ‘frame mark’ on late 18th and early 19th century paintings has been set out in section 1.2 above. To take two ‘three-quarters’ portraits in the National Portrait Gallery cut from narrow rolls, we learn from the frame mark that both were taken from canvas of roll length 17 yards. But while Lemuel Abbott’s Viscount Bridport, 1785, came from a roll marked as 75 hundredths of a yard, i.e. 27 ins wide (fig.2), Henry Perronet Briggs’s Charles Kemble, c.1830, came from a still narrower roll marked as 70 hundredths, i.e. about 25 ¼ ins. The long side of both canvases was evidently taken from the length of the roll, rather than across.

The three-quarters can be found by other names. Many 18th-century artists avoided the term when dealing with clients, preferring to call it a head, as did Sir Joshua Reynolds and Sir William Beechey. [81] It was sometimes described as a bust or bust-length. [82] Very occasionally it was called a quarter-length as in the Lely Executor Accounts in the 1680s and in John Singleton Copley’s American correspondence where the system he used in 1771 of quarter, half and whole-lengths was probably influenced by his habit of charging ‘two busts to a half Length’.[83] Since the mid-19th century the three-quarters has generally been described as a half-length. [84]

The three-quarters size allowed a figure to be portrayed almost to the waist at life size. Few artists reckoned to include the hands of their sitter in a portrait of this format partly because of the constricted size, but also due to the work it involved. J.S. Copley in 1771 was charging extra for children’s portraits in this format ‘because of the addition of hands’. [85] Joseph Wright of Derby in about 1790 charged 15 guineas, rather than his usual 12, for ‘A 3qrs of Dr Darwin with a hand’ and 17 guineas for ‘Mrs Biscoe 3qr with hands’. [86] In 1809 Sir William Beechey was asking 5 guineas on top of his usual 40 guineas ‘for a 3qr with a hand’. [87] Such portraits are exceptional because most artists regarded the three-quarters format as too restricted to include the hands.

Kit-cat format. Godfrey Kneller’s 2nd Earl of Godolphin, c.1710-2, 35 3/4 x 27 3/4 ins (90.8 x 70.5 cm), a portrait from the Kit-cat Club series (National Portrait Gallery).

Fig.8 Kit-cat format. Godfrey Kneller’s 2nd Earl of Godolphin, c.1710-2, 35 3/4 x 27 3/4 ins (90.8 x 70.5 cm), a portrait from the Kit-cat Club series (National Portrait Gallery).

2.3 The kit-cat

In the second half of the 17th century the next standard size up from a three-quarters was the very much larger half-length which was two and two-thirds times as large in painted surface area. [88] It was the need for a format which allowed the artist the flexibility to include one or even both hands comfortably in a life-size portrait to the waist, without having to paint a canvas of half-length size, which led to the development of the kit-cat size, 36 x 28 ins (91.5 x 71 cm), in the late 17th century, a format which was only a third larger in area than the traditional three-quarters (fig.8). [89] The kit-cat takes its name from the portraits of the members of the Kit-cat Club painted by Sir Godfrey Kneller between 1697 and 1721 and in turn the club owed its name to the mutton pies (‘kit-cats’) provided by Christopher (or Kit) Cat whose tavern near Temple Bar in London was the first home of this celebrated Whig dining club. All forty finished Kit-cat Club portraits in this format in the collection of the National Portrait Gallery include one of the sitter’s hands. It is sometimes said that the room in which the portraits were originally hung was too low to take the larger format of a half-length, but there is no contemporary evidence to support this idea. [90]

In his study of the Kit-cat Club portraits, J. Douglas Stewart identified three reasons for the new size, two aesthetic and one practical. [91] Most notably, he suggested that by the late 1690s there seems to have developed a ‘predilection for height at the expense of breadth’, to be seen, for example, in the lengthening of spoons and chairbacks, and that Kneller himself followed this trend, perhaps influencing the design of the kit-cat size, with its taller proportions. Interesting as this idea is, it is not susceptible to verification. Stewart also identified a possible desire for greater realism: the kit-cat size allowed the sitter to be presented on the scale of life (unlike a 30 x 25 ins canvas showing the same portion of the body). As a result, it meant that practically the scale of the head could be standardised for all sizes of portraits without the need for scaling up or down. However, the evidence is lacking as to whether artists actually experienced the difficulties that this analysis supposes. Stewart saw Kneller’s inspiration for the kit-cat size as deriving from the Holland, specifically from his master, Rembrandt, and his circle, and more remotely from portraits such as Raphael’s Baldassare Castiglione. [92]

In England portraits of approximate size 36 x 28 ins are uncommon before Kneller. In the case of Mary Beale her husband records purchasing three yards of sacking on 25 March 1681 to make three ‘yard clothes’, but this size is rarely found in her work. This was the same ‘excellent Royston sacking’ that Charles Beale had bought a month before to make three-quarter canvases, as already mentioned. While the three-quarters would have been set across the roll, the yard canvases would need to have come from the length of the roll. [93]

Beale’s practice in this instance of taking the long side of his yard canvases out of the length of the roll may not have been the norm judging from the conservation records for four portraits at the National Portrait Gallery. Herman Verelst’s John Locke, painted in 1689, an oval of 35 ½ x 29 ¾ ins, has selvedges at top and bottom (with very large and uneven tacking margins) and Kneller’s Francis Godolphin, 2nd Earl Godolphin, a Kit-cat Club portrait, appears to have been taken from the end of a roll of perhaps a yard width judging from the cusping marks. [94] More than a century later canvas rolls of about this width were available to George Hayter when painting his portrait of Lord Lynedoch in 1832 and to Thomas Phillips when painting John Dalton in 1835, sizes 35 ½ x 27 ½ ins and 36 x 28 1/8 ins respectively; both canvases retain selvedges at top and bottom. [95] All four portraits were thus taken from rolls of width at least 36 ins, and perhaps 37 or 38 ins wide.

Fig.9 Vertue’s format diagrams. George Vertue’s diagrammatic indications of different formats, which he devised ‘to describe instantly at view the size, magnitude or form of any picture – especially for portraiture’. Top row, ‘head or portrait 3 qrs’, ‘kit-cat’, ‘half length’ and ‘doubble ½ len.’ Second row, ‘whole at length’, followed by less often used sizes (British Library, Add. MS 22042, repr. from Vertue’s Notebooks, vol.5, p.113).

Fig.9 Vertue’s format diagrams. George Vertue’s diagrammatic indications of different formats, which he devised ‘to describe instantly at view the size, magnitude or form of any picture – especially for portraiture’. Top row, ‘head or portrait 3 qrs’, ‘kit-cat’, ‘half length’ and ‘double ½ len.’ Second row, ‘whole at length’, followed by less often used sizes (British Library, Add. MS 22042, repr. from Vertue’s Notebooks, vol.5, p.113).

Already within Kneller’s lifetime the kit-cat was being adopted as a general term for a particular canvas format. It was so used by the young John Smibert in 1722 to describe one of the first portraits entered in his account book, a portrait of Henry Justice, as ‘Kit-Cat’, and it occurs in George Vertue’s notebook the following year to describe a self-portrait of Kneller himself as ‘a Kit cat with a Cap on’. [96] In about 1734 Vertue included the kit-cat in his notation system for recording canvases (fig.9). The new canvas size was used for subject pictures as well as portraits, generally by turning the canvas horizontally, for instance by William Hogarth in his Marriage à la Mode series in 1743, as James Northcote later thought worth commenting, ‘the size of those pictures of Hogarth of marriage alamode is an exact Kitt Cat canvas’. [97]

The kit-cat size, popular though it became, was not taken up universally. John Smibert, Joseph Highmore, George Romney and Thomas Phillips used it quite often, Allan Ramsay, Joshua Reynolds and Thomas Lawrence sometimes, while Thomas Hudson, Francis Cotes and Thomas Gainsborough rarely. Later, Francis Grant claimed, ‘I never recommend Kit Kat - unless it is to be a companion picture - It is too large for a head - & too small for the figure...’. [98]

2.4 The whole-length or full-length

The usual term for a portrait showing the whole figure was a ‘whole-length’ but ‘full-length’ can be found; it was used by Thomas Wright in his catalogue of Lord Derby’s pictures of about 1730, by Joseph Wright of Derby in his notebook in about 1760 and by Lady Holland in a letter to her sister in 1763. [99] Thomas Gainsborough used both terms in the course of a single letter to a friend in 1764, suggesting that they were interchangeable. [100]

Large portraits of this kind are more likely than others to have been cut down or folded back to reduce their size. Often their dimensions as quoted in exhibition and sale catalogues have been inaccurately measured or are sight sizes. Nevertheless, it is possible to come to some conclusions about the eventual standardisation of the whole-length at 94 x 58 ins (249 x 147.3 cm).

For much of the 17th century a rather smaller size was used, perhaps as a result of constraints in the width of canvas then readily available. James Northcote writing in the 1780s noted that, ‘Vandykes whole lengths is seven feet by 4 feet 2 in’, i.e. 84 x 50 ins (213 x 127 cm), and this size, or close to it, is indeed commonly found in Van Dyck’s English whole-lengths. [101]

In the case of a slightly earlier Flemish artist in London, Paul van Somer, his few whole-length portraits were about the same height as Van Dyck’s, within a few inches either way, but slightly narrower at about 47 ins, a common enough width at the time. [102] In 1618, when asked to paint a whole-length to a very narrow format, he thought that the breadth was ‘unfitting’, finding it ‘so exceeding narrow and unsuitable to the the other length [the height] that it is most unfitt and almost impossible for any man to squeeze up a picture in so narrow a rome…’[103] In practical terms, then, Van Somer had a clear idea of what size was appropriate for a whole-length portrait.

Fig.10 Whole-length format. Godfrey Kneller’s 2nd Viscount Townshend, 1704?, 94 x 57 3/4 ins (238.7 x 146.7 cm) (National Portrait Gallery).

Fig.10 Whole-length format. Godfrey Kneller’s 2nd Viscount Townshend, 1704?, 94 x 57 3/4 ins (238.7 x 146.7 cm) (National Portrait Gallery).

While Peter Lely’s whole-lengths are somewhat larger than Van Dyck’s, it is in Kneller’s mature work of the 1690s and early 1700s that the whole-length size of 94 x 58 ins, or within an inch or two, becomes the norm, as for example in Charles Townshend, 2nd Viscount Townshend, 1704? (fig.10). Kneller used a slightly smaller size of 92 x 57 ins for his Hampton Court Beauties, perhaps owing to their particular situation in the Water Gallery at Hampton Court, and occasionally he adopted a larger size of about 97 x 60 ½ ins. [104] Of Kneller’s chief rivals, Michael Dahl used approximately the same standard size of 94 x 58 ins in most of his whole-lengths and John Closterman in many. [105] It would seem that by the end of the 17th century the taste internationally for grander portraits, perhaps taken with an increase in the width of readily available canvas, encouraged the adoption of a larger size for whole-lengths.

A whole-length canvas size of 94 x 58 ins is found in the work of most early 18th-century artists including Jonathan Richardson, Charles Jervas, Joseph Highmore, William Hogarth and Allan Ramsay. Thomas Hudson sometimes adopted the narrower size of 94 x 54 ins. Both Sir Joshua Reynolds and Joseph Wright of Derby quote 7ft 10ins by 4ft 10ins (i.e. 94 x 58 ins) as standard, as does the author of The Artist’s Repository in about 1785. [106] A conspicuous exception in this roll-call of artists is Thomas Gainsborough, whose whole-lengths tend to be slightly shorter in relationship to their width, typically measuring 88 x 59 ins, 90 x 60 ins and 92 x 60 ins. [107] Some of Romney’s whole-lengths are also on the short size.

Very few whole-lengths survive unlined with selvedges intact. The National Portrait Gallery collection includes two unlined portraits relevant to the discussion. George Romney’s William Paley, 1789, showing his subject to the knees on a canvas size 59 7/8 x 48 ins, was cut from the sort of roll used for whole-lengths. The canvas bears a ‘frame mark’ indicating a roll width of 61 ¼ ins, which would have been suitable for a whole-length. After stretching, the canvas measures about 62 3/8 ins wide as shown by the selvedges at top and bottom. In fact, this canvas was painted to a format peculiar to Romney. It was recorded in his studio ledger as ‘H-W-L’, that is, a half whole length, meaning that it was painted on half of a whole length canvas turned sideways. [108] Sir Francis Grant’s Sir James Hope Grant, c.1861, a seated whole-length portrait of the artist’s brother, size 84 1/8 x 52 ½ ins, retains selvedges at left and right, with turnovers of 3 ¾ ins at left and 6 ¾ ins at right, indicating a roll width after stretching of some 63 ins. [109] Such a width was equal to seven-quarters, while a width of 62 ins featured in Winsor & Newton’s trade catalogue of c.1857-61.

In another instance, George Morland’s Inside of a Stable, exhibited in 1791 (Tate), the nominal canvas roll width can be worked out from the duty ‘frame mark’ as 165 hundredths of a yard, i.e. 59.4 ins. The picture actually measures 58 ½ x 80 ¼ ins, suggesting that it may have been painted on canvas from a slightly narrower roll than the paintings by Romney and Grant but still sufficiently wide for a standard whole-length.

A canvas of size 94 x 58 ins, with a height-to-width ratio of 1.621, is very close in proportion to the golden mean, or golden section as sometimes termed, of 1.618 but there is no evidence to suggest that this canvas size was chosen with the golden mean in mind or, indeed, was ever subsequently linked to the golden mean by those painting or viewing whole-length canvases of this size.

The size could be varied for special needs: portraits of the king or queen were often done on an extra large canvas while seated portraits and children’s portraits would be painted on a shorter canvas. An example is Thomas Hudson’s seated whole-length of Lord Dumfries, which is an unusual size at 74 x 51 ins, selected specifically to fit the overmantel in the drawing room at Dumfries House in 1755. [110] In the late 18th century a further even larger standard whole-length canvas was introduced, as is discussed in section 3.2.


Found a mistake? Have some extra information? Please contact Jacob Simon at jsimon@npg.org.uk.