Beccari
Closer study of the two Beccari treatments of the Black Sea, in comparison with his surviving chart of 1403, might perhaps reveal that, besides introducing a large body of new place-names and adjusting the mismatched scales of the Atlantic and Mediterranean, Beccari may also have introduced subtle but significant improvements to the coastal outlines of the Black Sea. Compare, for example, one small feature, the island at the mouth of the river besides faxio/fasso [47] at the mid-point of the sea's east coast. The upper Cornaro version reproduces precisely the old-fashioned 'spanner' shape of the 1403 chart but the lower version is different. On the other hand, on the 1403 chart the oval shape of the island beside the Danube delta and Fidonixi beyond that (looking somewhat like the number 2) are very close to the lower of the two Cornaro versions (the one with Beccari's name). By contrast, the upper version gives the delta island a circular shape while Fidonixi is trifoliate. [See further about faxio/fasso on The features chosen to illustrate the varied portrayal of islands and estuaries on portolan charts up to 1469.]

A check of the eleven Black Sea names first otherwise seen on the 1403 chart found those repeated on each Cornaro version. Only palormo (Table of 'Significant Names' (a Microsoft Word document) – see no.365b) and bogancim (=bogoman, no. 370) are missing. On the other hand bosaro (presumably boscam, no.403) first noted on later charts, but not that of 1403, appears on each of the Cornaro versions.

The proven veracity of the toponymy and outlines on the Cornaro copies provides confidence in these mirror images of two lost charts by Francesco Beccari. Though which is the earlier, and how each relates chronologically to the 1403 original, is unclear. Unfortunately, the two charts by Francesco Beccari's son Batista do not include the Black Sea, so it is not possible to check what would have been Francesco's mature version of that area.

Benincasa
In the same way, two surprising features were noted in the shape of Corfu on the two Benincasa charts. Pujades's Cor 10 shows twin bays set into the north-west coast – a feature not found on any of the other Benincasa charts available in reproduction – as well as a distinctive arrow-head shape to the promontory projecting north-east. The latter, which is repeated on Cor 9, is also unusual but can be seen on the earliest surviving Benincasa chart, that with the date 1461 (Pujades C 62). Unfortunately, the reading is unclear on the other Benincasa chart in Florence assigned to that same year (C 63). From 1463 until at least 1469 (when the Pujades scans cease) Benincasa shows, consistently, a rounded promontory instead of the arrow-head.

If we are to trust the Cornaro Atlas, and the evidence above suggests that we should, the two-bay and arrow-head form for Corfu on Cor 10, followed by the arrow-head alone (as found only on the original of 1461 and Cor 9) indicate that its draftsman was using two models from the start of Benincasa's career, one of them earlier than any survivor. This is considerable weight to hang from one island shape and perhaps others will investigate this point further.

What did the Cornaro copyist add or not trouble to reproduce? For some reason he imposes his own consistent, and very distinctive style on Rhodes, with a large red Maltese cross over a plain background. However, comparison of the shapes he gives the island with those on the equivalent surviving originals shows that, as elsewhere, he has faithfully reproduced the distinctive outlines in each case.

Rhodes neatly sums up the Cornaro scribe's approach. He was recording, with care, the coastal and island configurations of the original, and its toponymy (even if there are some noticeable mis-readings). He was not, though, concerned with the chartmakers' visual signature. Although he uses the diagonal corner scales introduced by Benincasa, regardless of their authorship, he does not otherwise imitate any of the visual conventions that distinguish the work of his models from one another. As a result, some of the unattributed work in the Cornaro atlas remains to be identified – a process that will require careful comparison of toponymy and coastal configurations.

Where the Cornaro atlas is an unreliable witness is in such trivial, but diagnotically useful features, such as the detached 'a' for Aigues Mortes [23] and the 'd' or 'da' for Damiata [51]. The copyist's disregard for the original island and estuary colouring shows that he preferred to apply his own colours rather than reproduce what was in front of him. Indeed, as can be seen by comparing the Cornaro version of a Roselli chart with that author's invariable practice, the Venetian scribe has also supplied colour where the original had omitted it. For these reasons, the analysis of the relevant Cornaro sheets has not been incorporated into the general C&SA results, since it would unhelpfully confuse the visual signatures that can be discerned there. However, one of the C&SA tables is devoted to the Cornaro Atlas (a Microsoft Word document).

The other respect in which the Cornaro Atlas copies are not true facsimiles concerns their arrangement on the page. Having decided to place the reproductions onto the uniformly-sized folios of a volume, whose spread was considerably less than that of the original loose charts, the copyist was forced to split each into two or three. There is no guarantee, either, that the sheet-lines of the atlas pages were preserved. Whether the copying was done directly, e.g. by tracing, thus preserving the original scales, remains to be checked. { This paragraph added 5 December 2014}

What must have been a significant collection of original portolan charts, mostly but not all Venetian, and spanning perhaps 80 years, was presumably in one place so that Pesina (?) could made his copies. There is no indication that the models had been damaged by shipboard use – as far as I know there are no gaps in the coverage – and it seems more likely that this collection of model charts would have been the property of a mercantile or patrician owner. As such, its chances of survival would have been higher than that for an individual, working chart. If they weren't destroyed, is there the tantalising possibility that they might still be preserved somewhere in Venice? {This paragraph added 30 November 2014}

Finally, as it appears that none of the Cornaro charts was, coincidentally, copied from any of the originals that survive, as far as coastal outlines and place-names are concerned, it gives us, effectively, an extra eight works by named 15th-century chartmakers, with parts of three others.

Virga [23]
(for the detailed Colour & Shape Analysis findings see here – a Microsoft Word document)

[23]. Venice, Correr, Port.40
Pujades C 28, Venetian, beginning 15th, see 2007, p.495a
Campbell 121a

The suggested attribution by Pujades presumably rests on toponymic similarities between this and the signed Virga chart (although the two sequences are certainly not identical). In other respects, there is little to indicate Albertin de Virga as the author of this work. Neither this nor the 1409 chart is clearly enough reproduced to be able to look for handwriting similarities but the visual signatures of the two works are markedly dissimilar. The portrayal here of the Danube delta [41-2] as an arrangement of six separate 'islands' is unique. The eastern Canaries – included here but unfortunately not on the signed chart – are noticeably crude. Perhaps most significant of all is the absence here of the detachment and westwards shift of the Black Sea, seen uniquely on the 1409 chart. So, unless, a handwriting comparison from better scans provides supporting evidence, or the markedly dissimilar handling of the C&SA elements between the two works can be argued as evidence that one represents an early and another a later production, I do not think this attribution is sustainable. Any toponymic similarities identified by Pujades are more likely to point to a similar Venetian source. With those provisos, I conclude that this work is by an otherwise unknown chartmaker.

At a conference in 2011 (published in 2013), Pujades provided convincing evidence for the attribution of the Venice chart to Virga (2013a p.35 (note 54)). He kindly provided me with comparative details of the handwriting on the signed and attributed charts. It is quite clear that both are in the same hand and I am more than happy to admit my error. {This paragraph added October 2013}.

Italian: early 15th century [24-5]
(for the detailed Colour & Shape Analysis findings see here – a Microsoft Word document)

[24]. Oxford, Bodleian, Douce 390
Pujades A 13, Italian, beginning 15th
Campbell 57

No attribution has been suggested for this work, whose place of origin is also in doubt. There are sufficient peculiarities in its visual signature to indicate strongly that it is by an otherwise unknown chartmaker. Examples are the almost exclusive use of blue and green, and the colour and labelling of Majorca (different on sheets 4 & 5). The shapes of Ceffalonia and Zakynthos [see Explanatory notes and wider implications (Zakynthos)] are reminiscent of the unusual forms found on the 1367-83 Pizzigano atlas. Although their place-name sequences are not close, it would be worth comparing this further with the Pizzigano.

[25]. Milan, Ambrosiana, F.260 Inf.(2)
Pujades C 31, Venetian, c.1420
Campbell 85

This has a few distinctive visual features, for example the treatment of Euboea ( negroponte ) and Cyprus. White strips left across the width of each island breaks their blue colouring into strips. The large armorials on Corsica, Sardinia and Sicily, two circular and one square, are unique [see The presence of an armorial device within the separate countries of the British Isles, as well as within Corsica, Sardinia and Sicily.] For Sardinia and Sicily, large white spaces have been left within the overall wash for name labels. This and the now lost (?) Kamal chart[No.19 above] are the only Venetian works to adopt that Catalan device. The application of an overall wash to Corsica and the two large Italian islands is noted on just one other 15th-century Italian chart (Pujades C 50) but the treatment there is quite different. This is evidently the work of another unknown chartmaker.

Cesanis [26-7]
(for the detailed Colour & Shape Analysis findings see here – a Microsoft Word document)

[26]. Genoa, Bib. Berio, 'Luxoro Atlas', identified as the work of Francesco de Cesanis
Pujades A 14, pre-1421
Campbell 81

[27]. Venice, Museo Storico Navale, ms 1749
Pujades A 15, Venetian, first quarter 15th [single sheet from an atlas covering the central Mediterranean]
Campbell 122

That the Luxoro Atlas is in the hand of the person who signed the 1421 chart, Francesco de Cesanis (Pujades C 32), is evident from various distinguishing features of its hand, notably the wavy letter 'l'. It is instructive, therefore, to see just how much their visual signatures do, or do not match. About two-thirds of the colours are different on the two works, thus demonstrating that Cesanis was one of those chartmakers who did not use a standard palette. However, about the same proportion of the noted conventions – name labels, island shapes, etc – are the same. The similarities apply particularly to the Ionian and Aegean islands. For many chartmakers colour choice was of little importance; the hydrographical details, and particularly those that would have formed part of the outlines traced from the model, will naturally tend towards consistency.

The reason for linking here the Luxoro Atlas with the single surviving sheet of the Museo Storico Navale atlas is because of a single feature. The Museo's sheet extends to the west just as far as the Rhône delta and the first name is pinea. The pi (rather than just p) is written on the large island used to denote the delta next to vignom (Avignon). That in itself is strange but this is close to the point where many charts, Venetian among them, place the red initial 'a' of the name Aigues Mortes [23], and this seems clearly to imitate that convention. Since the separated 'a' was one of the diagnostic points I examined I can be fairly confident that, although not then looking for it, I would have noticed the detached 'p' or 'pi' of pinea nearby.

pinea refers, now and certainly in the 18th century, to a pinewood near to Saintes Maries de la mer. {This sentence added March 2015, on the basis of a personal communication from Jacques Mille}

The Luxoro Atlas is the only other instance seen of pinea, written in this detached way. The name also appears, undivided, on two works with a common authorship, the Corbitis and Pinelli-Walckenaer atlases [21 & 22 above]. The name has not been noted other than on these four anonymous Venetian works, though, as it was not considered a 'significant name', its presence was not systematically checked. The sequence of names on the four unsigned atlases and the Cesanis chart is set out on the accompanying Table 2 (which is a Microsoft Word document).

The way that Cesanis, in the Luxoro Atlas, had separated out the first (detached) letter of pinea so that it aligns with Arles suggests that he was copying – not fully comprehending (although the 'p' is correctly in black against the red for Arles) – from an earlier logical arrangement (as on the Storico Navale, which is considered to be somewhat earlier). The word stagnom nearby is another non-standard name found, it seems, on some Venetian works of this period (for example Pujades A 29-31), though that point was not thoroughly checked. Kretschmer (1909, p.589b) records stagnom on Benincasa but does not list pinea at all. The Storico Navale atlas sheet, while linked to the Cesanis environment through pinea, has its own distinctive feature: decorative border sections that divide up the scale markings, comprising alternating arrangement of dots as on a tile from a game of dominoes, a green 4 and a red 5.

On Cesanis see also the next section.

Venetian: second quarter 15th century [28-32]
(for the detailed Colour & Shape Analysis findings see here – a Microsoft Word document)

[28]. Rovigo, Bib. Com., Silv. 182
Pujades A 29, Venetian, second quarter 15th; related to A 30 below
Campbell 106

[29]. Parma, Bib.Pal, II.32.1624
Pujades A 30, Venetian, second quarter 15th; related to A 29 above
Campbell 103

[30]. Vatican, Vat.Lat.9015
Pujades A 31, Venetian, second quarter 15th; same hand as the following – see 2007, p.496a
Campell 158

[31]. Venice, Marciana, It.VI.212
Pujades A 16(b), Venetian, second quarter 15th [the later sheet for the Adriatic (no.6) in the 1426 Ziroldi atlas]; same hand as the preceding - see 2007, p.496a
Campbell 116

[32]. London, British Library, Add.19510
Pujades A 12(b), Pinelli-Walckenaer atlas, Venetian (F. Cesanis, c.1434) – one of the two later sheets, that covering the Aegean (no.7)
Campbell 47

The first group (Pujades A 29-31) is distinguished by the frequency and number of its 'labels' – where space in an island's colour is left white to enable it to take a place-name later (or created by working round an existing name). The regular geometric shape given to Guernsey (in the English Channel) is repeated on all three (and found also on the 1421 Cesanis chart, the Luxoro atlas and perhaps elsewhere – though this point was not thoroughly checked).

Some unusual features also relate one or more of this group to other Venetian works. Examples are the serrated circle for the island off the south-west of Ireland ( montorio/brazil) [9] on A 30 and A 31, which can be seen also on the Cesanis chart and again later on. Multiple labels, e.g. on Majorca and the Aegean islands, link A 29 and A 30, and to a lesser extent, A 31. A label on Corfu is found only on A 29-31 and on A 16(b), the Adriatic chart added to the 1426 Ziroldi atlas; while labels on Chios and Skyros can be seen on A 29-31 but, this time, additionally, on one of the two later charts added to the Pinelli-Walckenaer atlas (A 12(b)).

Pujades suggested that A 29 and A 30 were 'related'. The visual analysis strongly supports that, although they are evidently not in the same hand – note the many distinctive characteristics of A 29, e.g. the vertical line projecting up in front of the 'a'. Pujades also identified the Adriatic chart added to the Ziroldi atlas (A 16(b)) as being in the same hand as A 31.

Might the overlapping similarities within this group, combined with some differences and the fact that only two (A 31 and A 16(b)) appear to be by the same person, provide evidence for the existence of some form of co-operative workshops, as distinct from a hierarchical atelier [see Venetian practice]?

Pujades suggested Francesco Cesanis as the author of the two later sheets in the Pinelli-Walckenaer atlas. Since no C&SA features were selected for the Adriatic above Corfu, only the Aegean could be considered here. There was insufficient evidence to support a Cesanis connection on those grounds alone.

Ziroldi [33-6]
(for the detailed Colour & Shape Analysis findings see here – a Microsoft Word document)

[33]. Milan, Ambrosiana, SP 2.39
Pujades A 21, c.1446
Campbell 88

[34]. Chicago, Newberry, Ayer 2
Pujades A 22, c.1446
Campbell 138

[35]. Vatican, Rossi. 676
Pujades A 24, second quarter 15th
Campbell 156

[36]. London, British Library, Add 18665
Pujades A 27, second quarter 15th
Campbell 46

Four unsigned atlases were confidently assigned to Ziroldi by Pujades, because their toponymy matched the chartmaker's signed work and because – at least for the last three – they repeat his distinctive treatment of what are presumably intended for inland lakes and rivers, presented like a string of kites. Some of these perhaps relate to the orcinoli, or circles representing the river's source, that E. Solopova finds characteristic of medieval maps [see Imago Mundi 64:2 p.156 (2012)]. {This sentence added 6 October 2013}

In addition, to this untrained eye, all the atlases appear to be in the same distinctive hand, in other words by Ziroldi himself rather than workshop productions (unless it is to be supposed that assistants were required to imitate all the master's scribal peculiarities). Note, for example, the use of a slanting dash to denote the dot on a 'i', and the way that some initial letters (G and S particularly) would be continued across the word, to link up with a double 'l'. Satallia (on the south coast of Asia Minor) is a good example. Given without any flourish in 1422, the subsequent elaboration of its initial letter, already evident in 1426, can be seen to its full extent in the signed works of 1443 and 1447. The four unsigned works show the same treatment for that, and other features.

The C&SA fully supports those attributions. Although Ziroldi was often inconsistent in his use of colour, his treatment of the analysed conventions was more predictable. One of his trade-marks (otherwise found only on the 1430 Briaticho and 1448 Nicolai atlases) was to split the usual oval shape for the imaginary island of scurçe (off the south-west of Scotland) [2] in two, with a larger left side offering a jagged edge to the smaller right one.

As a result, those four attributed works fit neatly into the chronological slots suggested by Pujades. Indeed, there are features in the C&SA which independently corroborate the Pujades sequence of Ziroldi works, although too much weight should not be given to such unimportant, and possibly accidental, elements, particularly as they comprise an ABA pattern where, following change, the original treatment returns.

But the suggestion that Pujades A 21 and A 22 fit into the period 1443-6 is supported, first, by the omission of the separated 'a' for aguemorte [23] and, strangely, for the inclusion of the separated 'd' for damiata [51], second, for Ziroldi's preference for an elaborate pattern of islands in the Danube delta [41-2] for those same years, and, third, for his convention, in that same period, of splitting the two large islands in the Dnipro estuary into three [45-6]. That same feature supports the late dating for A 24 and A 27, since the (now single) large island to the west is provided with a name label. One unusual feature, involving the atlas that Pujades puts as Ziroldi's latest work (A 27), is the small detail of a circle (or more strictly a regular hexagon) on Tenerife (inferno) representing Teide [14], but the handwriting does not cast any doubt that this is Ziroldi's work.

Italian: second quarter 15th century [37-42]
(for the detailed Colour & Shape Analysis findings see here – a Microsoft Word document)

[37]. Florence, Bib.Medicea-Laur., Gaddi Rel.9, 'Medici Atlas'
Pujades A 25(a) & (b), Italian, second quarter 15th, supposedly based, in part, on a lost work of c.1351, with later sources for larger-scale sheets covering the Adriatic and the Aegean (3 & 7)
Campbell 76

[38]. Venice, Marciana, It.IV.493
Pujades A 26, Venetian, second quarter 15th [same hand as the following] – see 2007, p.496a
Campbell 114

[39]. Florence, Arch. di Stato, C.N.11
Pujades, C 49, 'Italian', second quarter 15th [does not specify Venetian although identifies it as in the same hand as the preceding] – see 2007, p.496b
Campbell 72

[40]. Barcelona, Arch. Corona de Aragón, MP1
Pujades C 50, 'Italian', second quarter 15th
Campbell 127

[41]. Istanbul, Topkapi Sarayi Museum, 1827 [no scan available and so this could not be considered]
Pujades C 56, 'Italian', mid-15th
Campbell 134

[42]. Lyons, Bib de la ville, MS 179
Pujades A 28, Venetian, second quarter 15th – see 2007, p.496a
Campbell 8

The Medici Atlas is a confusing and unusual work. Its calendar starts in 1351 and Pujades speculates that much of the volume (that part which forms a single portolan chart spread over sheets 4-6 & 8) is based on a Genoese work of about that date. The toponymy is certainly 14th century, and Dulcetian rather than Venetian. However, Pujades dates the construction of the volume as a whole to the second quarter of the next century, and notes that it is all in the same hand even if some of the content reproduces originals of different dates. He concluded (p.504 note 116) that the atlas was produced, probably somewhere other than Genoa or Venice, 'based on previous material, including some of the works by Francesco Cesanis'. The likelihood of a Cesanis connection is discussed below.

Even if most of the Medici Atlas's charts were modelled on an assumed Genoese work of around the middle (or perhaps the second half) of the 14th century, a similar problem is experienced here as with the Cornaro Atlas. In each case the copyist will not have reproduced the 'visual signature' of the original chartmaker. Did the draftsman, however, faithfully recreate the hydrographical outlines? And, in the case of early Genoese works, what might those have looked like?

Early Genoese characteristics
Pujades has identified four charts and an atlas that he considers to be Genoese works from the first half of the 14th century [see above Genoese: first quarter 14th century (and the section after that)]. Nothing else has survived by Genoese practitioners until the Beccaris in the following century. It is instructive to look in particular at the treatments of the Isle of Man and Limnos on early Genoese productions. Even if the dating of the Beccari works might still allow them to be considered as models for the Medici Atlas, their outlines for those two features would rule them out.

The Isle of Man [3] was portrayed as a cross by the great majority of chartmakers, from 1321 until the 16th century. The exceptions were mostly Catalan, starting with a simple rectangular form on the 1330 Dalorto/Dulceti. This was elaborated, from the Catalan Atlas onwards, with symmetrical half circles set into the sides, and chamfered corners. This form was still found on the work of the Viladesters. The Medici Atlas outline is a simple progression from the original Dalorto rectangle, with a small groove at the bottom, like the gateway to a castle. Pujades A 26, the second work in that early Genoese group, breaks up that symmetrical form. Unfortunately the chart (C 49), considered to be in the same hand as A 26, does not include the Isle of Man. Finally, in the list of non-cross outlines, the 1424 Pizzigano chart has an idiosyncratic form of its own.

Of the five early Genoese works identified by Pujades only two include the Isle of Man. The distinctively different Paris atlas (A 9) has a unique representation, comprising eight angular headlands, but the outline on the Paris chart (C 11 – in so far as it can be discerned) although very different might perhaps be considered a model that had been simplified for the Medici Atlas.

The Aegean islands have been omitted from the Medici Atlas's general chart of the eastern Mediterranean (sheet 6). This was done intentionally, since the island outlines would have been part of an early drafting stage. Perhaps this exclusion related to the presence in the atlas of a larger-scale sheet devoted to the Aegean, based on a different model. Such a direct link between the construction of sheets based on 14th- and 15th-century outlines respectively demonstrates that the atlas, for its charts at least, is the result of a single overall plan. However, it does mean that there is no opportunity to compare on the two versions the island that is the best overall indicator for shape analysis, namely Limnos.

Instead, we must turn to the Ionian Islands, which occur here just once (on one of the 'original' sheets, no. 6). Comparing the Medici Atlas with those five early Genoese works we find a number of generally similar outlines, particularly on the two closely related charts (Pujades C 10 & 11). However the shapes for Keffalonia and Zakynthos could not have come from such sources. Most indicative is the Medici's Keffalonia, in which the curving northern headland that was already visible on the 1311 Vesconte chart, as well as on the oldest surviving Genoese work (the chart in the Riccardiana (Pujades C 4)), has here been turned into a full coat-hanger loop, as the culmination of a shape quite unlike any potential model identified.

It has, I think, been generally assumed that the content of the standard portolan chart sequence in the Medici Atlas (sheets 4-6 & 8) was based on a mid-14th century model, whereas the enlarged Adriatic and Aegean (sheets 3 & 7) represented later outlines (possibly, as suggested by the toponymic analysis, contemporary with the atlas's construction). If that was the case, any contribution by Cesanis (the person Pujades considers to have provided some of the models for this atlas) could apply only to sheets 3 and 7, or perhaps to some of the other general and non-standard sheets.

A Cesanis connection?
The Medici Atlas outline for Limnos [36] (on the later, larger-scale Aegean sheet) is certainly as precise and elaborate as those on both the 1421 Cesanis chart and the Luxoro Atlas (which is certainly in Cesanis's hand), but they must have different origins. The Medici version moves closer towards a true picture of this immensely complicated coastline than perhaps any other 14th- or 15th-century form. For example, it has twin bays in the north coast and, from west to east, two small and a single very deep bay into the south coast. By contrast, the 1421 Cesanis outline (found also, at least, on the Luxoro atlas, the Alvise de Cesanis copy in the Cornaro atlas, Pujades A 29 & 31, and, simplified, on A 30) has a single bay into the north coast, and a deep bay followed by a lesser inlet along the south. The fact that the 'Cesanis' outline for Limnos (inferior to that on the Medici Atlas) was in common use among Venetian practitioners of his time, and that Francesco's son Alvise (who was still alive in 1496, see Pujades, 2007 p.487) was repeating that same outline years later, is evidence of its being a standard representation, which Francesco de Cesanis would also have used on any model he might have provided for the Medici Atlas.

In connection again with Francesco de Cesanis's son Alvise, it is worth drawing attention to Guernsey (one of the British Channel Islands). The Medici Atlas's distinctive shape (like a curving piece of pipe) has a line of dots (rocks?) to its west, where the later copy in the Cornaro Atlas of an Alvise chart shows a related form, like a horizontal pin-cushion, with the pins to the north. Neither of those forms is at all like the symmetrical jagged outline found on Venetian work in the first half of the 15th century.

The use of a terminal full-stop after each place-name is a distinctive feature of the Medici Atlas and although it is also found on the Luxoro Atlas and Cesanis chart (and apparently nowhere else) there is no question of the Medici Atlas being in Cesanis's hand (on this see Pujades, 2007 p.495a and note 116 on p.504).

Far more noticeable, and evidently unique to the Medici Atlas, is the array of jagged islands north of Ireland. In presentation, these are similar to the treatment on the enlarged chart of the Aegean (no.7) but no equivalent model has been found. It seems unlikely that they were invented by the Medici copyist.

The tentative conclusion, on the basis of the Colour & Shape Analysis, is that, if the Medici Atlas is indeed based on a lost Genoese original, this would have been significantly different from any of the five surviving productions from that city that are supposed to date from before the Medici's presumed model of c.1351. Likewise, any connection with Francesco de Cesanis (rather than, more generally, with Venetian work of his time) would need supporting evidence.

Pujades has identified the Marciana atlas (A 26) and the Florence chart (C 49) as being in the same hand. I was not able to identify this unknown chartmaker's visual signature because of the poor condition of the atlas and the chart's restriction to the eastern half of the normal area.

The chart in the Barcelona Archives (C 50), however, has sufficient unique features to show that we are dealing with the single work of an unknown chartmaker. Most obvious is the treatment of the larger islands, Corsica, Sardinia, Sicily, Euboea, Crete and Cyprus, as well as the major estuaries. Each appears to be edged in gold, enclosing a full colour wash, into which a pattern of circles has been set (the scan is not clear). While a little reminiscent of the unsigned and unattributed Ambrosiana chart (C 31, to which Pujades assigns a date of about 1420), the treatment is quite different.

Two specific peculiarities further mark out the Barcelona chart from all other surviving productions. First, an arrangement of four dots to represent the Isle of Man [3]; second, a two-colour circular button, to represent the mythical man [10] (off the south-west of Ireland).

The Lyons atlas looks superficially like the work of Ziroldi with its corner-pieces, though these are emblamatic where those on the 1426 atlas mostly feature saints. That aside, there are no other obvious similarities. This atlas has two distinctive visual features. First, it is unusual in that two colours only are used for the islands, red and green. Second, it is one of three anonymous Italian atlases to supply Chios with a large red Genoese cross of St George [38]. The other two, the [21] Corbitis and [22] Pinelli-Walckenaer atlases, are by the same draftsman, whose distinctive hand (see Chapter p. 402) is quite unlike this. The only other similar treatment of Chios is on the charts of the Catalan Vallseca.

Benincasa [43-5]
(for the detailed Colour & Shape Analysis findings see here – a Microsoft Word document)

[43]. Paris, BnF D21815
Pujades C 74, Venetian, 'Benincasa atelier'
Campbell 24

[44]. Vicenza, Bertoliana ms 524
Pujades C 75, 'Italian', 'Benincasa atelier'
Campbell 124

[45]. Parma, II,29,1621
Pujades A 42, Venetian, third quarter 15th
Campbell 101

The Colour & Shape Analysis table devoted to Benincasa, with another for his successors, relates closely to the separate essay on that prolific chartmaker: The style and content of Grazioso Benincasa's charts: imitation, innovation and repetition

Here, it is sufficient to comment on the remarkable consistency in the C&SA profile across Benincasa's productions and the conclusion that all are evidently in his hand or that of his son Andrea. I could see no evidence in favour of a workshop operation, and the two partial charts that Pujades attributed to a hypothetical Benincasan atelier (his C 74 and C 75) are better interpreted as fully-fledged works by Benincasa from which their original signatures (which would have been at the western neck) had been trimmed off. [On Benincasa's consistency see further Explanatory notes and wider implications (Consistency: Benincasa).]

For Pujades there was no doubt that Benincasa 'was a specialised craftsman, for not even the most ardent sceptics deny that such regular and systematic characteristics needed the support of a well organised atelier employing permanent collaborators' (2007, p.497a). One of the comments I made in the 1987 Chapter (p. 431a) lent accidental weight to that: 'Occasional clumsiness – for example, several attempts at scraping the hidden circles on sheets of one of the British Library's Grazioso Benincasa atlases (Add. MS. 6390) or an abandoned circle on the 1424 [Pizzigano] chart – suggests the inexperienced hand of an apprentice'. In fact, the inexperience seems to have been my own since a new look at charts 7 & 8 on the Pujades DVD (his A 39) shows that the jagged lines are not part of any circle but, while presumably made by careless use of a pair of compasses (dividers), would instead have been the fault of the atlas's owner – thus almost certainly providing additional evidence of the use of such volumes at sea. [The comment about the 1424 chart, where there are actually several false circles, does however remain valid.]

The Parma atlas (Pujades A 42) is a close copy of Benincasa's and is also considered in the essay on that chartmaker – Benincasa, his successors and imitators.

Vallseca [46-7]
(for the detailed Colour & Shape Analysis findings see here – a Microsoft Word document)

[46]. Florence, BN Centrale, Port.16
Pujades C 41, Vallseca atelier, c.1440 – see Pujades, 2009, p.332, 336
Campbell 78

[47]. Paris, BnF, Cartes et Plans, D3005 (six fragments)
Pujades C 43, Vallseca atelier, c.1447 – may be in the same hand as the chart dated 1447, see Pujades, 2009, p.332
Campbell 22

Summary of conclusions
These are tentative. Pujades's claim that the two charts above are products of a Vallseca atelier is broadly supported by the findings of the Colour & Shape Analysis. But those results, and a closer examination of some island forms, suggest that the charts from this atelier, whether signed or not, reveal considerable variation between the productions of what appear to be several different people. Part of that seems to result from Vallseca's coastline model excluding those medium-sized islands that other chartmakers might have traced (or, more likely, pounced). But, in general, if all these charts were produced under Vallseca's direction, it must have been a very loose one.

The putative Gabriel de Vallseca workshop is the most interesting of those proposed by Pujades, since the justification for it is developed in his 2009 study of Vallseca's 1439 chart (pp.327-30). Vallseca signed a contract to produce 24 charts over a period of half a year, in other words at a rate of one a week. Extrapolating from that, Pujades estimated that Vallseca might have produced as many as 2000 charts over a lifetime, a quantity that would have necessitated the semi mass-production procedures of a workshop. However, for a different interpretation see above, The Vallseca contract of 1433.

Because of the price quoted for an individual chart, Pujades is confident that the 1433 contract concerned utilitarian, non-embellished charts. There is nothing that might correspond with those among Vallsecan survivors. Since the C&SA concerns three signed and dated charts, as well as the two discussed here attributed to a Vallseca atelier, and since all those have typical, and expensive, Catalan adornments, it might seem unlikely that the C& SA findings could throw any new light on possible workshop practice relating to the simplest kind of chart. However, would not all Vallseca charts have been produced in a single workshop, with most of the procedures shared, since the ornate chart is no more than an embellished version of an ordinary one? [See What is the difference between a plain and a de luxe chart?.]

Some observations, though, can be made. Pujades (2007, p.493a) had noted two different hands on the 1439 Vallseca chart, and of the five surviving works attributable to the Vallseca 'firm' (signed or unsigned) four appear to me to be in different hands. Pujades identified two works at the Bibliothèque national de France, one dated 1447 and the fragmentary pieces (his C 43 – which could originally have had a signature) as the work of a single draftsman. They are certainly closely similar though I noticed that the 1447 chart (like the signed chart of 1449) used a very plain initial 'C', whereas the other three works add a curling extension to the top.

Different people working from the same basic model is what would be expected from workshop practice. But the visual analysis throws doubt on that. Not because of different treatment of ornamental features, since those might have been the work of an artist perhaps drawing freehand and introducing intentional variations. That the arms placed centrally over Sardinia and Sicily are inside a diamond on the 1439 chart, within a circular frame on the attributed Florence chart (C 41), and presented like a shield on the 1447 Paris chart and its supposed (fragmentary) copy, while being absent altogether from the chart of 1449, is probably of no great significance. What matters is that such special treatment for those islands seems to have been a part of Vallseca's visual signature. [On that see further The presence of an armorial device within the separate countries of the British Isles, as well as within Corsica, Sardinia and Sicily]. Similarly, C 41's lack of the Genoese cross over Chios, found on the three signed charts [but not covered on C 43's fragments], need not cast doubt on the Vallseca authorship.

However, what raises interesting questions is the variety of island outlines found for the studied areas in the Ionian Islands and Aegean archipelago. It is probably true to say that most chartmakers repeated broadly, or exactly, the same outlines from one chart to the next. Indeed, since the outline on their pattern might differ from other practitioners, it can be a way of recognising an individual's work. With Vallseca this would be very difficult. It is hard to make the point without illustration but compare, for example, the 1439 and later Vallseca outlines for Kefallonia and Zakynthos (Ionian Is) as well as Limnos and Skyros (Aegean), adding in those on the attributed Florence chart of c.1440 (C 41). The obvious conclusion is that these were drawn freehand and carelessly on the basis of a common model. Since the draftsmen were presumably aware that the outlines of Limnos and Skyros, at least, were wholly imaginary, this would not have mattered. But this seems to indicate that, in Vallseca's studio, those larger islands did not form part of the traced or pounced outlines, in contrast, for example, to the precisely reproduced configuration given to the Nile delta. If so, this would seem to differentiate Vallseca from the practice adopted by other practitioners who appear to have transferred those larger island outlines to the new chart along with the continental coastlines [though a resolution of that point requires access to comparative images].

ymador
Since the C&SA does not suggest any reasons to challenge Pujades's attributions, there are other hydrographical observations that raise further questions about the way that the Vallseca atelier operated. The first is the large blue island of ymador to the south of the Canaries found on the attributed Florence chart (C 41, with a suggested date of c.1440) but not on the signed Vallseca chart of 1439 (although the island might have been considered to lie beyond the limits of that work). Cortesão (1969-71, 2, p.151) notes that this island is reminiscent of the himadoro on the 1424 Pizzigano chart. But why, if the Florence chart is later than that of 1439, was ymador not included on the latter? Surely it would have been present on the workshop model, unless that island was added to the model after 1439. The two later Vallseca charts cannot help here as they do not extend beyond the Mediterranean.

faxio/fasso
Another hydrographical point does raise questions about the model used for the construction of the Florence chart (C 41). It concerns faxio/fasso [47] at the mid-point of the Black Sea's east coast, marking the point that a major river (the Enguri) emerges, with a variously shaped island at its entrance. On most charts this river, whose identity was obviously not clear to the makers, has to be assumed (and that includes the Catalan/Estense world map) but Vallseca, at least on the 1439 and c.1440 Florence charts, actually depicts the river. Besides that, there are major differences between the treatments on those two charts.

In 1439, the small red island is placed, logically, at the river entrance, alongside ffasso and between lipotimo and paliostoma, with san grigo as the next name to the south. This is the pattern found earlier on the three Viladesters charts (1413-28). The Florence chart has no fasso and places the red circle three names further to the north, well separated from the large river, which is here placed south of s. grigo, i.e. two names further south than in 1439. The 1447 chart is truncated at that point but the related fragments in Paris and the 1449 chart are similar and evidently show a different later conception of that area. fasso reappears in its earlier position, though there is no river (at least on the 1449 chart) and the island now takes on the 'spanner- head' shape. The difference between the 1439 and Florence charts may, once again, highlight the distinction between what was copied from the model (the main coastline) and what was added freehand (the placement of the island and river). But, assuming that the Florence chart was an official production of a Vallseca workshop, it does not indicate close oversight by the person in charge.

R. Soler [48]
(for the detailed Colour & Shape Analysis findings see here – a Microsoft Word document)

[48]. Paris, BnF, Cartes et Plans, B8268
Pujades C 51, second quarter 15th, attributed to a Soler atelier – see Pujades, 2009 p.313
Campbell 16

[Berlin, Humboldt Univ., H14-12
Pujades C 52, second quarter 15th, signed R. Soler but any date there might have been is now illegible
Campbell 34]

Pujades (2007) p. 504, note 87 managed to read the name 'Rafel Soler', but not a date, on the vellum neck of the Berlin chart. This was enough to move one chart from the anonymous to the signed list. He also attributed a Paris chart to R. Soler and provided comparative illustrations of the two on p.259 (see also pp. 492-3 for a discussion of the two works).

The signed work stops at the end of the Mediterranean, which deprives us of useful Colour & Shape diagnostic features in the Atlantic. But the match of tones and treatments is sufficient to show that Rafel Soler was unusually consistent even if – a very small detail this – he chose, on the signed work, to place on the island the first two letters of Aigues Mortes (a unique feature of that chart) [23] and likewise the 'da' of Damiata [51] in the nearby island, while the anonymous work just separates out the first letter in each case.

The condition of neither chart is particularly good and it is difficult to be certain about the letter forms. But are they not both in the same hand, which leans to the left? The treatment of the initial 'I' of Inglatera on the unsigned Paris chart [the signed work does not include the British Isles] is highly distinctive and illustrative of Soler's occasional flourishes. More useful for comparative purposes is the contraction for the first three letters of Tripoli (both in Lebanon and Libya). The latter name, Tripoli de Barbaria, can be seen most elaborately on the signed Berlin chart. If it can be confirmed that these two charts are in the same hand, is there any need to posit a Soler workshop? Can we not say instead that both are by this chartmaker, whom Pujades has rescued from obscurity?

Pujades suggests (2009, pp.314-5) that Rafel was continuing the tradition of his grandfather Guillem Soler, of whom we have two works datable to the period 1368-85. However, comparison of the C&SA features from the work of Guillem and Rafel show little similarity, which is perhaps not surprising after a gap of 50 or so years.

Rafel's work not only has one or two distinctive features, at least two of his conventions seem to have been imitated by Benincasa:

• A four-leaf clover on Tenerife [14] – next seen on Benincasa
• The two stripes on Majorca, arranged SW/NE [20] – next seen on Benincasa
• Detaching the first two letters of Aigues Mortes ('ay') [23] – unique
• A diamond shape for fassio / faxio [47] - unique

Roselli [49-51a]
(for the detailed Colour & Shape Analysis findings see here – a Microsoft Word document)

[49]. Paris, BnF, Cartes et Plans C 59096
Pujades C 70, third quarter 15th, Roselli workshop
Campbell 20

[50]. New Haven, Yale University, Beinecke Library, 49cea/1425
Pujades C 71, third quarter 15th, Roselli workshop
Campbell 145
[NB because the scan on the DVD is not of high quality, use instead the Beinecke Library's webpage. Click 'See all images', then select 'Image ID: 1027161', finally click on the 'Zoom' option to bring up the MrSID image.]

[51]. Modena, Estense, C.G.A.5b
Pujades C 72, third quarter 15th, Roselli workshop
Campbell 93

[51a]. Modena, Estense, C.G.A.1 'Catalan (Estense) world map'
Pujades (2007) did not discuss this
Campbell (not included in the Census but see Chapter p.379 (note 71)) [for a detailed note, see the separate entry below]

[--- The fragment of a chart with Ludwig Rosenthal (Munich), Catalogue 163 & 167 (1920s?) was attributed to Roselli, but could not be considered here.]

[--- New York, Hispanic Society, K15]
Pujades C 73. As explained in a pop-up note when you click on that image on his DVD, this was mistakenly considered to be a fourth Roselli workshop production, but once Pujades saw a reproduction, he realised it was Venetian and from the later 15th (or more probably early 16th) century. The tables on pp.65 & 206 of his 2007 book include erroneous entries for this chart, which could not then be corrected. It should be pointed out that the Hispanic Society had described it as late 15th-century Venetian, see Sider (1992) p.8.]

On the details of Roselli's style and development see Heinrich Winter, 'Petrus Roselli', Imago Mundi 9 (1952): 1-11. Having completed the note below I looked at that article and noticed that Winter discussed the Modena chart (on p.11), noting that the border trimming had removed the author's legend but preserved the 'typical ornamental moulding ... so that his authorship may be regarded as certain'. Winter also mentioned the Paris and New Haven maps, concluding: 'It is true that there are some similarities to the manner of Roselli, for instance the city images, but they are only similar, not identical with those designed by Roselli, which are the same on all his maps'. So, independently, we came to the same conclusions.

Summary of the conclusions
Paris (C 70) was not drawn by him, nor does it imitate his work.

Yale (C 71) is visually closer to Roselli's style, but deviates from his invariable treatment. The unusual form of the initial 'R', found on this and on the Paris chart, suggests a possible common authorship.

Modena (C 72) is apparently in the same hand as Roselli's signed work (presumably his own) and its overall appearance matches that of the fuller charts Roselli produced from 1464 onwards, This should therefore be treated as the eleventh surviving work by Roselli, having originally been signed on the trimmed western edge.

The Catalan/Estense world map's probable attribution to Roselli is confirmed and a date of c.1462-4 proposed [see the separate entry below].

The results of a survey of the large and small visual features of Roselli's productions finds no evidence pointing to the existence of an atelier, in the sense of productions created under the direction of the master chartmaker but distinct from his own work.

Ten signed charts by Roselli have survived, two of which are not available in high quality reproduction. One has the same date as his earliest chart of 1447 and was at one time with Kenneth Nebenzahl in Chicago. It was illustrated in his Catalogue 20, item 164 with a photograph which is sufficient to discern the larger visual features only (Campbell Census 178). The other is from 1469, a later date than any surviving chart. It was originally with Vollbehr and then sold at Sotheby's in 1988 but no illustration is available to me (Campbell Census 180 and now E.20). Of the other early chartmakers, only Benincasa has left us more signed works.

The eight available scans give a good corpus against which to compare the three works attributed to Roselli. First, we need to consider the signed charts. What can we learn from those about the consistency of Roselli's island and estuary colouring, the shapes of significant features, and any visual development that might have taken place over the 21 years spanned by the accessible surviving charts?

A glance at the Roselli Colour & Shape Analysis (and ignoring the three attributed works) is sufficient to demonstrate that, among the signed works, there is a great measure of predictability in the use of tints, particularly since the mauve/red/pink distinctions are probably due only to my fallible perception or differences in ageing. In four cases, the earliest accessible chart (that of 1447 in Volterra) is different from all those that follow. Given the consistency afterwards, this presumably represents a conscious change by Roselli between 1447 and 1449 coupled, perhaps, by an appreciation of the advantages of consistency. This seems to conflict with Pujades's statement (2007, p.493a) about Roselli's 'certain stylistic and chromatic diversity, although invariably within the confines of a characteristic style clearly distinguishable from that of other cartographers'. If, however, that comment refers just to the larger visual elements, it would serve to emphasise my contention that the smaller features analysed in the C&SA tables tend to take their consistency from being part of the basic unthinking construction, whereas the ornament - presumably drawn free-hand by an experienced illuminator – would inevitably vary, possibly intentionally.

Limnos
However, if the colours were applied consistently, the same did not apply to the shape of the Aegean island Limnos (stalimene) [36]. The island's contorted outlines prompted a wide variety of diagnostically useful shapes from different chartmakers [on which see Explanatory notes and wider implications ('Limnos')] and Roselli's charts show erratic variation. It must be stressed that the attempts at Limnos's shape, some of which are clearly fanciful, were not just casually drawn freehand but conform, more or less faithfully, to one from a range of patterns going back to Vesconte. Since extracts cannot be taken from the DVD accompanying the 2007 Pujades volume, I cannot illustrate this point, and describing a complex shape is not likely to be helpful. Nor have I been able to compare the variant shapes but have had, instead, to remember them or draw each one out. The analysis crudely counts the number of 'lollipop projections', emerging potentially from three of the four sides of the island, but a small change could turn a lollipop into a narrow finger promontory – an alteration, which since it dealt with imaginary geography, is not worth overstressing.

But, to generalise, Roselli started in 1449 (the 1447 chart culminated at the Adriatic) with an outline for Limnos [36] close to that found on the 1403 Francesco Beccari chart. Then, with his next surviving production, the chart of 1456, he moved to a markedly different outline, and one not observed elsewhere, with a single large bay into the north coast and a pair of deep bays in from the south. Since that was an attempt at realism it is surprising that he added two 'lollipops' projecting to the west. Thereafter Roselli reverted, at least in 1462 and 1466, to his original version. However, in 1464, 1465 and 1468, he used a variant of the Beccarian form, which transformed the two westward 'lollipops' into promontories. So, if the original form is A, the unique 1456 form B, and the amended original C, the sequence of seven charts runs thus: A, B, A, C, C, A, C. We can only speculate what that might tell us about the multiplicity of patterns owned by Roselli and why different ones were used.

Before looking at the three supposed atelier productions, it is worth testing if any developments can be discerned among the more obvious visual features of Roselli's charts, since these could have bearing on the dating of the attributed works, assuming they are indeed based on Roselli originals. Viewed strictly chronologically, there are seven elements that do not appear before a given date but are regularly present thereafter. Of course, any of these supposed 'developments' might be no more than coincidence. Since they are all additions rather than changes, they could easily have been omitted afterwards.

Visual features
From the outset, Roselli's charts include city vignettes and many flags, as well as the large green Granada mountain. His early style can be seen in the chart of 1447 in Volterra – as confirmed by the reduced illustration in the Nebenzahl catalogue featuring the other chart with that date. By 1456, Roselli was using a wavy border decoration; by 1462 he had placed armorial shields on Sardinia and Sicily (they had been there on the three British islands from the beginning), and he added (usually) five wind heads around the edge; until finally, by 1464, he had extended to the north to take in the Baltic, and inserted inland rivers, the Atlas Mountains and African rulers in their tents.

It had been hoped to use two other features that were introduced in 1464, namely a picture of the Ottoman sultan inside his flag-topped tent in Asia Minor, and a vignette of Fez with its flag. However, those disappeared after 1465 and 1466 respectively, so their absence from the Paris and Modena charts (the one in Yale does not cover either area) cannot be considered significant. Nor, perhaps, should much be made of the fact that the Modena chart is unique in including vignettes for Barcelona and Valencia (to go with their flags, which are found on all Roselli charts) as well as the fourth in a line of identically flagged vignettes just south of the Baltic and next to the horseshoe of the Bohemian mountains and the vignette of Prague. This one is apparently labelled coconia, which name is repeated nearby. As carconia (?) it appears, without its flag, on the Catalan Estense world map [see below, where there are further comments on Roselli's stylistic development].

It might have been supposed that the level of decoration would have corresponded to a differential price structure, as Pujades surmises in the case of Vallseca (2009, p.329). But, with six separate elements displaying the identical chronological pattern of absence until a particular date and invariable presence after it (with perhaps the earliest five charts lacking the feature and at least the last four having it) it must surely be the case that these 'macro' visual features reveal a chronological development on the part of Roselli. Unless an attempt is made to argue that the more elaborate the chart, the more likely it is to survive, thus skewing the record.

The only possible exception to what is, I am aware, an unfashionable developmental interpretation relates to the lack of the wavy border on the 1465 London chart. However, the vellum appears to have been trimmed around the edges, perhaps taking with it any such border there might have been.

A Roselli atelier?
It is now time to turn to the three charts suggested by Pujades as the product of a hypothetical Roselli atelier. None provides the complete coverage of a full chart. The Paris chart (Pujades C 70) goes no further east than Sicily, the Yale chart excludes the Atlantic, and the trimmed Modena chart (the fullest of the three) now ends before the Canary Islands. The linked Table 3 (which is a Microsoft Word document) relates the colour match of each attributed work to what was normally (or exceptionally) found on a signed Roselli chart. A similar comparison is made with the work of some of his contemporaries.

The three works proposed as examples of a hypothetical Roselli workshop can now be assessed against the C&SA findings and the development that can be seen in the chronological sequence of his surviving productions. A comparison of the handwriting is attempted, and certain peculiarities in the attributed charts are taken into account. The findings of the statistical analysis (Table 3), plotting the match of island and estuary colours and shapes, is borne out in relation to the steadily growing complexity of the visual adornment of Roselli's charts.

The linked Table 4 (a Microsoft Word document) plots the relevant features as they appeared on the signed and attributed charts. All three attributed works should be placed after 1462 because of their inclusion of features, missing from the first five signed Roselli charts drawn up to that year, and then invariably shown afterwards. If based on his models, these three charts must therefore date from the later part of the period of Roselli's known production or even beyond it.

The Paris chart (Pujades C 70) is the furthest from any Roselli model here, just as it was in the Colour & Shape Analysis, though allowance has to be made for the fact that it covers just the Atlantic coasts and then as far as Sicily. Because of its peculiarities, it cannot be interpreted as a simplified version of a Roselli chart. There are some colour choices in the Atlantic not found on any signed Roselli work but the most obvious peculiarity of the Paris chart are its heavily lettered names for Ireland and England, which take the space of the shields for the three countries (with Scotland) found on all Roselli charts. Note particularly the highly ornate initial letter for England.

The Beinecke Library chart in Yale University (Pujades C 71) has a closer C&SA profile (though not nearly so close as that with the Bertran & Ripoll chart) but again there are several colour choices that do not appear on any original Roselli work, among them a blue Corfu and a mauve Lefkada. Limnos is again useful here, since its outline is one I have not noticed elsewhere. The uncoloured Majorca (not discernible on the scan supplied to Pujades but visible on the higher resolution Beinecke scan (see above)) is noticeably different from the Roselli model, which would (after 1447) place about six diagonal red stripes (north-west to south-east) over underlying gold (or possibly silver).

The third attributed chart, in Modena (Pujades C 72), offers a perfect match in its colour choices to the Roselli norm (leaving aside three illegible features). However, its Limnos is slightly different from the shape on any signed work. Although broadly similar to the outline found on the chart of 1449, it has been modified to leave just one 'lollipop' along the west coast. It is reasonable to consider this as just another Roselli variant of the Limnos outline. Overall, this chart, with its extension to take in the Baltic, is very similar to those produced by Roselli between 1464 and 1468.

Handwriting
Now, the question of handwriting. Others with the appropriate expertise may disagree with me but I can see no reason to doubt that all the signed charts are in the same hand, presumably that of Roselli himself. This has few obvious peculiarities but the curling extensions from the top of the C and G are noticeable (and also found on some of the Vallseca charts and Bertran & Ripoll's of 1456). These features are not present in the attributed Paris and Yale charts, which instead have a detached dash where the signed charts have the extension. Those two works also share a very distinctive initial R, of which each chart has more than one example. See, for example, on the Paris chart, Redondela and Ribadeo (Portugal) and, on the Yale chart, Roma and Rasacalxero (at the point where the north African coast turns to the north). If that suggests the same hand was responsible for the Paris and Yale charts (or might that device have been used more generally in Majorca at that time?) then the noticeable differences described above need to be explained. The simplest answer would be that the same reproductive scribe was copying from different models – but that depends on confirmation that just one hand was involved.

Unfortunately the Modena chart image on the Pujades DVD is not sharp enough for confident analysis but it seems to me to be in a hand very similar if not identical to Roselli's. That would certainly fit in with the chart's perfect match in the C&SA. Direct visual comparison with all the other signed charts is not currently possible, but if it were to become feasible in the future it might well confirm the strong affinity already revealed through comparing the illustration of the 1464 chart in the Pujades volume (pp.324-5) with the DVD's scan of the Modena chart (C 72). The Modena chart looks in every way like a Roselli production.

The Modena chart is complete along the top but trimmed on the other three sides. By analogy with complete signed charts, and working from traces left to the east and south, it is clear that it would originally have had the wind heads, town vignettes and tented African rulers that formed an invariable part of Roselli's visual style after 1464. Significantly, the truncation to the west passes exactly through a supplementary intersection point, itself an extension (along the central east-west line) from the most westerly of the main intersection points. All the surviving charts show that Roselli placed his signature to the west of that subsidiary intersection point (or in just one case, level with it). Had there been a signature on the Modena chart it would almost certainly have been trimmed off. No other clearly Rosellian chart is unsigned. If that was the case here it would be unique among the ten survivors, and there is no obvious reason why that should be so. There is a strong likelihood, therefore, that this chart, rather than being a 'workshop' production, was, in its full original form, another signed and dated work by, and in the hand of, Roselli himself. The three additional features noted above could perhaps indicate that it should be dated later than 1468. Pujades notes that Roselli was active up to 1489 (2009, p.487).

As for the Paris and Yale charts, the evidence set out above makes it hard to support the claim that they were produced in a Roselli workshop. If there is a particular character to the Rosellian toponymy, which led Pujades to attribute these works to his hypothetical workshop, then some common or indirect source must have provided that information. The creator of the Paris chart made no attempt to imitate Roselli's style – indeed he flaunts his own individuality – and, on the grounds of handwriting, island and estuary colouring, as well as the omission of elements that are invariably found on Roselli's authenticated productions, there seems no reason to treat that or the Yale chart as having been produced under the direct supervision of the master, Roselli. That is surely what a workshop must imply. Instead, if the shared capital R form found on both Paris and Yale charts is confirmed as the work of a single individual, he needs to be added to the list of hitherto unknown practitioners, even if as no more than a copyist.

Catalan Estense world map [51a]
(for the detailed Colour & Shape Analysis findings see here – a Microsoft Word document)

[51a]. Modena, Estense, C.G.A.1 'Catalan (Estense) world map'
Pujades (2007) did not discuss this
Campbell (not included in the Census but see 'Chapter' p.379 (note 71))

For an enlarged scan see the MEDEA-Chart database.

In the 1987 chapter referred to above I was over-hasty in mentioning this merely in a footnoted list of those world maps that incorporated portolan chart outlines, but whose 'scale was rarely sufficient for more than a sprinkling of names'. For details of other possible candidates, including two fragments of large world maps in the Topkapi Saray, Istanbul (for which Pujades could not obtain scans), see Destombes (1984, pp.193-234).

The so-called 'Catalan (or Estense) world map' displays the typical embellishment of Catalan portolan chartmaking. It has been assigned to the period c.1450-60 and, specifically, to the chartmaker Petrus Roselli (Pere Rossell). The questions of date and authorship will be considered in a moment.

There are two good facsimiles of this, the most recent (with an accompanying CD) being Antichi planisferi e portolani (Modena: Il Bulino; Milan: Y. Press, 2004). I have access instead to the facsimile published by Moleiro in 1996 in Barcelona, with a text by Ernesto Milano and full transcription of the place-names. Apart from a differently arranged rhumb line network, to take account of the much increased extent to the north, east and south, it is, at its heart, a portolan chart, and deserves to be considered as such in the literature, for its content rather than its purpose. Indeed, like the Catalan Atlas, it is actually more a world chart, than a world map. The treatment of the interior of the regions not normally found on a portolan chart is handled in a similar way to those parts of Europe and North Africa that are usually present on Catalan charts.

The normal portolan chart region occupies almost exactly half the overall diameter of the Estense world map, which, at 113 cm, is only a little larger than the east-west dimension of a typical chart at around 90 cm. This may well have been a factor in the choice of scale for this world chart. The dense toponymy of a portolan chart (Pujades recorded 1,854 names on the 1439 Vallseca chart) demanded a minimum scale. Below that it would be too difficult to write the place-names in the restricted space available. So, rather like zooming in or out of an Internet map today, a drop in scale would mean a large drop in the number of names. The world map drawn by the chartmaker Albertin de Virga is a good example of how the coastal names had, effectively, to be abandoned because of the work's modest scale. To have reduced the names by, say, a half, would have involved the mapmaker in a time-consuming process of selection. It would have been possible to include just the significant (red) names, as was done for a literary work, Dati's 'La sfera', but there is no evidence that any chartmaker ever did that. Otherwise the realistic choice was between the full usual coastal toponymy or none at all. [Which is the labour-saving route taken by fakers, see Fakes: Notes on selected maps – 'Portolan charts (1509-25)'.]

Sample checks reveal that the Estense map, despite being drawn at much less than the scale of a typical chart, still displays almost the full selection of expected place-names. Comparison with the list of Valencia coastal names given for Roselli (Pujades, 2007 p.397) shows that a small number of names have had to be sacrificed (including Alicante, invariably in red). This probably applies particularly in the congested areas around peninsulas, and would reflect logistical necessity rather than toponymic editing.

The reason for labouring this point is to establish that the Catalan world 'map' was certainly made by a Catalan chartmaker, and as an extension to a portolan chart rather than as a newly conceived work. The Catalan Atlas provides one of the precedents for this, as, on a more modest scale, do the 1339 Dulceti and 1439 Vallseca charts. While the Estense's extension gives much attention to the Indian Ocean, the join between the surveyed Mediterranean and Black seas on the one hand and the imagined coastlines beyond is clear evidence of the latter's source in literary texts rather than navigators' experience. The contrast between the Black Sea, left clear of shading, and the Caspian Sea, treated with coloured wavy lines, provides a graphic illustration of the boundary between the portolan chart and the world map. {This sentence added 11 June 2016}.

Earlier researchers thought this world chart belonged to the 14th century, but its inclusion of West African discoveries made by Alvaro Fernandes in 1446 established that as a terminus post quem. As R.A. Skelton acknowledged in his entry in Destombes (ed.) Mappemondes, (1964, pp.217-221, specifically p.221), it was Marcel Destombes in 1955 who had identified Roselli as the most likely author. Skelton himself found a close connection with that chartmaker's later work (1464-8). Does the Colour & Shape Analysis (C&SA) throw any light on the work's authorship and dating? Is it by Roselli and, if so, how does it fit into his stylistic development?

Toponymic selection and name forms are important tools in portolan chart analysis, just as commentary on the visual elements of illustrated Catalan charts needs an art historian's expertise. Neither of those elements, however, could be considered in detail here.

Authorship
If the Estense map is not by Roselli might any of his known Majorcan contemporaries be candidates? Apart from the possibility of it being the work of one of the shadowy figures identified by Pujades (2007, pp.486-7), we have surviving Catalan productions from that period by Rafel Soler, Gabriel de Vallseca and the two who combined on the 1456 chart, Jaume Bertran and Berenguer Ripoll.

The C&SA findings initially provide cautious confirmation of Roselli's authorship, with an 88% match [see Table 3 (which is a Microsoft Word document)]. However, when looked at more closely it is evident that none of the Estense colour choices is different from those found on Roselli's charts. Three of the four points of difference between the Estense map and the Roselli norm involve no more than a lack of colour where it would be expected – on the mythical Atlantic island of man [10], on the Aegean island of Lesbos [37] and on the large island in the Dnipro estuary [45] – while the fourth concerns the small estuary island, shaped like an old-fashioned spanner, at the mid-point of the Black Sea's east coast, near fasso/faxio [47]. Normally left uncoloured by Roselli, it is here coloured red, as by most other chartmakers. Omitting colour may have no significance, particularly when working at a smaller scale, and the absence of the island habitually found at the Danube entrance is probably due to that same cause. But adding a bold dash of red to a feature Roselli never otherwise noticed, is unexpected.

A similar exercise found a 48% match between the Estense map and the signed and attributed R. Soler charts, a 77% match with the single Bertran & Ripoll chart, and an 82% correlation with the combined work of Vallseca. Considering that I found a 100% match between Roselli's certain work and one of the charts attributed to him – coincidentally, also in the Estense Library in Modena – these findings cannot (on their own) be considered conclusive evidence of a Roselli authorship. Although Vicenç M. Rosselló i Verger, had noted in his 'Les cartes portolanes mallorquines' (2000, p.62) that there was little doubt that it should be attributed to Pere Rosselli around 1460.

Handwriting
The question of handwriting is for an expert palaeographer (which I am not) but it cannot be ignored. The Catalan Estense map is obviously not in Soler's distinctive handwriting [see comments above under R. Soler]. The variety of hands identified in the work of Vallseca or his atelier, make it difficult to identify such work. The latest surviving Vallseca chart is dated 1449 but, according to Pujades (2007, p.486), he was active as late as 1467, which could leave his workshop as a possible candidate for the Estense map. Finally, the Bertran & Ripoll scan on the Pujades DVD is not of sufficient quality to make out precise letter forms. The National Maritime Museum's online scan [search for 'bertran'], can no longer be enlarged to legible resolution [March 2012]. When viewed in the previous high resolution [2011], the hand (or hands?) seemed different to that of the Estense.

Thus, Destombes's attribution to Roselli remains highly probable. A number of the letter-forms in the Estense's handwriting match those in Roselli's work. A comparison of the name forms for the Valencia coast used by Roselli (from Pujades, 2007, p.397) with those on the Estense (as given in the 1996 facsimile) reveals considerable agreement – apart from the occasional omissions from the world map because of space restrictions. A similar comparison could be made for the much longer sequence covering most of the Adriatic.

Dating
On the basis of an assumed Roselli attribution, the further question of dating comes into play. The Estense map is conventionally described as c.1450 or c.1450-60. As already mentioned, it cannot be earlier than the west African discoveries of 1446, to which it refers. But how late might it be?

Skelton pointed to similarities between the Estense map and the later Roselli charts from 1464-68 [no image is available for the chart of 1469]. The Estense certainly includes several elements which appear on the last four accessible Roselli charts, having previously been omitted from the earlier five. Extending north to take in the Baltic involved a major decision to increase the range of his standard chart. However, adding inland rivers, rulers in North Africa and the Atlas Mountains were elaborations of what was already there. But do those help with the dating of the Estense map? Those features, and some others that come and go on the later Roselli charts, can all be seen on the 1439 Vallseca chart, and presumably on other lost works from that source. Why, then did Roselli wait 17 years to include them [1447 to 1464]? And if he had produced the Estense map in, say 1450 or 1455, with its inland rivers, Atlas Mountains and African rulers, why did he not also add those to the already much-embellished charts of 1456 and 1462?

West Africa
The year 1464 is significant for Roselli's work, in a different way. As discussed in the Benincasa study Africa – 'Stage 3', in that year (if not before) Roselli extended the west African coast to the south, giving it a new outline along with a sequence of fresh names. None of Roselli's earlier dated works (1447-62) had gone beyond Cape Bojador (buyetdor). For the coastline below that headland the Estense's names broadly match the earlier name sequence, introduced on the 1448 Bianco chart, and with none of the later names included. On the charts of 1464 and 1465 Roselli displayed instead the first part of the new toponymic list seen in full on the extra sheet added to Benincasa's atlases from 1468 onwards. Because of space restrictions on those charts, even with their extension to the south, the section of toponymy Roselli could show involved no more than a re-naming of part of the coastline discovered up to 1444. This allowed him to document the names a little beyond Cape Bianco, but that was well short of Andrea Bianco's terminus in 1448 (on a chart devoted to the Atlantic and arranged vertically rather than horizontally). [For the comparative name lists, including those for the Estense map, see Table 5: Comparing West African place names from samotamat to cauo de sancta maria (a Microsoft Word document).]

But space restriction did not apply to the Estense map, which, like the Bianco chart, goes as far as C. Rosso [here cgroso], with an unconvincing C.Verde next to it. Since Roselli must have come into possession of the new Portuguese information no later than 1464 (when he first includes it), there is no reason to suppose he would have reverted to the superseded outline and toponymy after that date, even if on his charts of 1466 and 1468 he chose, as before, to stop at Bojador.

As pointed out by Relaño (2001, p.103) there is a note beside C. Rosso indicating that this is the end of the western part of Africa. The coastline on the Estense map changes colour at this point, emphasising the distinction between the known and the unknown – further evidence that its author was not in possession of new information. {This paragraph added 26 January 2013}.

Indeed, the new African information might have come to Roselli a little before 1464. The names added by Benincasa in 1468 below C.Rosso (discovered by Fernandes in 1446 and the terminal point for both Andrea Bianco and the Estense World) come from the same explorer's voyage in 1447 and then, after a long pause, from Sintra's of 1460 and 1461. Since Bianco had not been aware of the Fernandes discoveries made the year before he drew his 1448 chart, and since such discoveries were not recorded on any known intervening chart, it seems likely that all the information on developments in the period 1447-61 would have reached Roselli at the same time, and via a Portuguese chart this time, not the hearsay Bianco had to depend on. If so, that would mean that Roselli saw that hypothetical chart between 1461 and 1464.

It might have followed logically from the link highlighted by Skelton between the Estense map and the more ornate style seen on Roselli's work from 1464 onwards that the Estense map would be assigned to the period 1464-8 (or, more strictly, 1462-8, from the date of the last of his less-elaborated charts). However, this seems not to have happened. Instead we have been presented with weakly supporting datings somewhere in the decade after 1450. Ernesto Milano's text accompanying the 1996 Moleiro facsimile offers a chronological resumé of earlier opinions. The Cape Verde Islands – not shown on the Estense map – feature prominently in a line of reasoning that attempts to date the map between the apparent discoveries of Cadamosto in 1454-6 and the islands' settlement by Antonio da Noli in 1460 (pp. 83, 123). Arguing from absence is always questionable.

Because of its west African content we can therefore be sure the Estense was created during the period 1446-1464. The visual evidence is less certain. The earliest accessible Roselli chart, that of 1447 in Volterra, treats at least four C&SA features differently to the way they were consistently handled afterwards. None of those appears in the same way on the Estense map.

Unless there are other justifications for the earlier provisional dating that means we can broaden the potential range slightly, from the conventional c.1450-60, to c.1447-64. This would take account of the possibility that Roselli could have included the pictorial elements, first seen on his work in 1464, on lost charts earlier than that of 1462 which omits them. But he is highly unlikely to have reverted to an archaic account of the Portuguese African discoveries, which are given such prominence on the Estense world map. With so much evidence pointing to the first part of the 1460s, my preferred date for the Estense map, therefore, would be c.1462-4.

Ramon Pujades (2023) has carried out a very thorough examination into the authorship and dating I had suggested and endorses my 1462-4 dating [p. 241/263]. However, he also brings highly detailed arguments to refute my assertion that certain unsigned charts should not be attributed to a Roselli [Rossell] atelier [pp . 244-/266 - 251/273]. I am happy to accept those. {This paragraph added 26 March 2024}

Arabic [52]
(for the detailed Colour & Shape Analysis findings see here – a Microsoft Word document)

[52]. Milan, Ambrosiana, SP2. 259
Pujades C 54, first half 15th
Campbell 89

The so-called Maghreb 'chart' extends from the British Isles to Corsica and Sardinia. It contains few Colour & Shape diagnostic features and little can be learnt from it in that respect. It is, however, unusual in using just green (?) and red for the island colouring.

For an extended note on this chart see the Census update page (no.89).