by Caroline Flick, translation by Susanne Meyer-Abich
The search functionality of the Getty Provenance Index (GPI) has become a standard tool for provenance research, which can yield unexpected successes when searching for further appearances of an object in art sale catalogues, yet can also lead to frustrating outcomes.* Since mid-2025 the new version GPI Remodeled was meant to ease data access, which implied that a closer inspection of the actual data was now obsolete. Nevertheless, for the Remodeled version data were transferred without additions or subtractions into a complex data model for Linked Open Data with subject-predicate-object information, based on the textual data of the legacy database, which also continues to be accessible for the time being as GPI Legacy.
Users can extrapolate the entire textual data in csv format â each of the thirteen files made available in GitHub contains around 15,000 entries.1 From this mass of data, the present article examines a selection from âsales contentsâ. Once again, the focus is on the Berlin auctioneer Hans W. Lange, because the business activities of this company in 1937-1945 offer a manageable number of objects for a relevant period of time, with previous research by the author permitting an assessment.
Neither the familiar intuitive search form in GPI Legacy nor the presentation of hits according to catalogue, artist, object as well as material and transfer allow an inference to the actual number of data fields in the respective dataset.2

The illustration shows the title line for csv files with all fields which constitute the complex structure of the datasets, based on the abovementioned subject groups. At the same time, the line illustrates a diagnosis quoted from one of the few reviews of the GPI Remodeled: The thin layer of data in the GPI would be hardly visible in the extensive framework.4
Taking a random example, it seems to become an illustration of this evaluation. The dataset illustrated below shows a lot of two watercolours offered in 1938. Its 18 entries in 503 fields suggest a number of queries with regard to using GPI Remodeled, but these are outside the scope of the present analysis.5

Our concern are structure and content of the datasets in GPI Legacy. The illustration first displays the data fields on the auction sale with numbers and dates, lot number, and page number, which were extracted with an object from the respective auction catalogues. These are followed by the auction house and by the full description of the respective lot.
However, this is where the problems start. The entry is quoted as follows âEin Paar Aquarelle. Römische Ansichten: St. Peter â Vesta-Tempel. Seide, 12 x 8,5 cm; 14,5 x 10 cm.â In the catalogue, the entry is separated from the previous lot by a section break.7 The traditional typography used in these catalogues adopts the dinkus to conclude a textual passage before starting another section, here with anonymous works.

It separates them from the previous entry, a signed watercolour âVor dem Pferdestartâ by Horace Vernet (1758â1836). Nevertheless, the pair of watercolours and even the three following lots are also attributed to Vernet in the datasets, because the dinkusâs meaning was not recognised. As a consequence, Vernet appears as âartistâ for these objects, including life dates, norm, nationality, and reference from the Union List of Artists (Ulan).
In the sequence, the description was then split into the data fields âobject typeâ, âmaterialsâ and âdimensionsâ. This leads to further errors: âwatercolourâ lacks material and the measurements of â12.5 x 8.5cmâ suggest a single object rather than a pair; silk as a significant search element is not transferred to the field category âmaterialâ. If a user inspected the two small pictures in the original, subject matter and terms of description would be more likely to yield results, that is a full-text search instead of a systematic one. The âtransactionâ is unknown because the price list only mentions sale prices over 100 RM and this one is missing.9 As the two watercolours were listed at 50 RM in the firmâs internal list of estimates (which is missing here) â it is conceivable that the sale price remained under 100 RM but the lot was still knocked down and sold.
The complex structure can therefore not be described as filled consistently or just evenly. The processing of reading, transferring, and correcting does not prevent such errors, which significantly impede, if not exclude, searching and potential identification. Users are well advised to bear these repercussions in mind, especially with view to searches by category.
The incorporation of these data in the GPI was achieved for the first set in 2013, based on the German Sales (GS) project of digitalisation of German auction catalogues in 1930â1945 by UniversitĂ€tsbibliothek Heidelberg, with the years 1900â1929 added in 2019.10. OCR text was corrected and structured, then cleaned âmanuallyâ and linked, and finally published in the GPI Sales catalog in 2013.11 At least for the collection areas represented in the GPI, searches are considered to be effective and reliable, compared to the much criticized full-text search in GS, which lists double or triple hits, or none. Such misinterpretations as these call this verdict into question.12
Data dictionaries for the Sales Contents provide explanations for the data fields and title abbreviations used to structure the information, listed under the respective names with field descriptions and examples up to a range of 140 fields. However, in an export of âfull recordsâ and examples for search results in GPI Legacy the csv file only offers 60 fields per object, because the fields appear much reduced and in part joined together. 13 A search result in Legacy is displayed as âflat fileâ, regardless of the underlying range of fields, in this instance the phantom works by Vernet.

Selection
In the 1980s, the GPI was intended as an instrument for the data of the Getty Trustâs own museum collection. The first collations appeared as books in print, later on compact disks, finally in a database and eventually on the web. Due to the museum interest, the focus was on painting and sculpture.
Hence, out of over 16,000 objects offered at auction by Hans W. Lange (1904â1945) in 35 sales, the csv files list 3,600 lines respectively lots.15 The substantial decorative arts share of the offers in such auction houses like furniture and textiles (at the firm in question 2,822 objects), porcelain and faĂŻence (3,335), small decorative items [Kleinkunst] (1,903), or Chinese and Japanese objects (1,691) remain invisible. But in this sample, a fifth of objects comprising paintings (2,735, with miniatures and drawings 3,154) is included almost entirely.16
Nevertheless, 2,735 paintings, with some pieces missing from the category, do not result in 3,600 datasets. Further data are produced for 41 miniatures (out of 143 counted for this firm), 261 drawings (276) as well as 91 small sculptures (129), 149 bronzes (232), and 361 sculptures (415), which makes for a somewhat rigid selection of the objects thus classified by the auctioneer.
The GPI therefore contains 93% of paintings and 78% of sculptures offered at this Berlin auction house. A comparative sample shows that for a large collection such as that of Margarethe Oppenheim (1857â1935), containing over 1,232 lots at auction at the firm of Julius Böhler in Munich in 1936, 79 objects are in the GPI, that is all 12 paintings and 67 out of 100 bronzes. With 100% and 67% respectively, a single owner sale has a similar rate of representation as the offering of the Berlin firm of auctioneers.17
The selection of objects does not follow the categories of the auctioneers but was clearly edited to include for example only 609 of the 776 sculptures offered by Lange. Architectural and objects for everyday use were excluded, not always consistently.18 For example, reliefs and cartouches are missing, but among antiquities, a Roman head of an amazon is included (400405/D-2400 0154), while a head of Dionysus (430127 275b) is excluded. Since the former came from a well-known collection and the latter is an early copy, this points to an art historical reflex action applied to text and perhaps illustration, but not at the object level.19
Consequently, French bronze putti (410925/D-2620 0203) from the collection of Jakob Goldschmidt though from âmetalworkâ [Metallarbeiten] are included, but the artist bronzes by Constantin Meunier and August Gaul from the collection of Max Cassirer (1857â1943) were missed, as they also appeared under âmetalworkâ (430127 482 f.) but had been included before when categorised as âsculptureâ [Plastik] (420513/D-2733 0315 f.). A âBull, stridingâ [Stier, schreitend] reappears as âbronze statuette of a bullâ [Bronze-Statuette eines Stieres] (400405/D-2400 0162) resulting in two GPI numbers for the same object. Similar pieces are treated differently, even when consigned again.
The inconsistent treatment (not only) of sculpture is again demonstrated by a chance find. An auction sale in 1930 was documented both in a regular sale catalogue and in a working copy â the latter without illustrations, reduced by around a third in text and typography, intended for bidder annotations during a sale â for the Camillo Castiglioni (1879â1957) collection, comprising 558 lots.20 Both catalogues were entered into the GPI, treated as separate events and are also listed with different total lot numbers â not an isolated case.21
The auction catalogue in the GPI (D-286) indicates 176 lots, the working copy (D-18) however has 167 lots, with just 154 matching. Based on the auction catalogue, a further 21 sculptures appear, the working copy has 10 different sculptures, two paintings, and an etching. Consequently, the âMarble Bust of an Emperorâ (420513 322) only appears once as a sale from the Castiglioni collection in 1930, that is solely in the auction catalogue (D-286 0072), Pietro Solaris: âEmperor Augustusâ (430127/D-2860 0319) however appears in both Castiglioni catalogues and therefore twice (D-286 und D-18 0102).22
Although paintings are easier to specify, a complete listing cannot be taken for granted, even if the share of listed objects is much higher at 93% than for sculpture. Gaps appeared where an auctioneer listed objects in an adjoining category when lacking the volume to make up a separate section, and the editors therefore eliminated them again. As an example, prints were sometimes listed under drawings, engravings (381119 218 ff.) also under paintings (390208 006 ff.).23 Occasionally, the auctioneer listed prints by the same artist under paintings for commercial reasons, so that underbidders could go for a lithograph instead: In the Alexander Prentzel (1875â1955) collection, five paintings by Franz KrĂŒger (1797â1857), were followed directly by two lithographs (420513/D-2732 0075 ff.), which do not appear in the GPI although one was even acquired by a museum.24
The editors of the painting section took a view on art historical quality based on descriptions and, ideally, illustrations. The âPortrait of a Bearded Warriorâ, according to Wilhelm von Bodeâs (1845â1929) expertise an authentic work by Rembrandt, only appears in a Graupe auction of 1934, but not in 1941, when the 1934 buyer Curt Elschner (1876â1963) resold the work in 1941.25 Since Elschnerâs paintings are otherwise listed in full, it seems unlikely that this one was overlooked.
Instead, the fact that an entire auction is missing suggests an error: In May 1941 Hans W. Lange offered 50 paintings as a separate group in advance of a large auction sale of decorative art, including several works of notable provenance which have been the subject of previous research.26
To begin with, the decision to include only a selection from the wide object range offered by art auctioneers was a logical outcome of the Getty Research Instituteâs focus on fine art, and it continued to be the case. However, inconsistencies inevitably restrict the searchability of the offers and therefore the records. When the editors generously included elements from one auction house or sale but (many) fewer ones from the next for reasons unknown, when an auction was missed or included twice, possible repeat sales, and consequently the movements on the art market, were not reliably recorded.
Object data
The 3,600 datasets which were entered follow the description in the catalogue, lot number, and date. This is the consistent base quantity.27 Yet in the field âartist infoâ, around 900 artist names remain without information, among them 767 anonymous works predominantly in sculpture.28
In the following, there are remarkable differences in allocating normative data to artists in accordance with the thesaurus of the Union List of Artist Names (Ulan).29 There are 2,550 sets, because 516 sculptures and 534 paintings are only attributed contextually, as in âItaly, 16th Centuryâ. There are 84 paintings in the catalogues where the artist cannot be linked to Ulan. For some, the âartist authorityâ was oddly set to ânewâ; others lack an existing Ulan and thus appear like rare and obscure German painters; further names were linked once but not when repeated , thereby some familiar or even well-known painters were excluded: Osias Beert d.J. (1622â1678), Eduard Wilhelm Pose (1812â1878), Otto Rethel (1822â1892) or Friedrich Gauermann (1807â1862) and Carl Blechen (1798â1840) â for mysterious reasons and resulting in the fact that a search for artists via standardised data is unreliable.30
The next group of fields concerns object materiality, with few options ranging from âmaterialsâ, âdimensionsâ to âformatâ, and also âgenreâ, âsubjectâ, and âinscriptionâ. The fields are filled in a rather varied manner.
Material as a category is essentially open to interpretation and will refer to the support such as panel or to working material such as oil, ink, or clay.31 Only 3,363 objects contain data on material, in 237 it is therefore missing. These do not necessarily suggest a materiality which could not be determined, instead this information was frequently not extrapolated, as in a âDoveâ made of gilt bronze (390622/D-2334 0084) or âTwelve Apostlesâ out of limewood (401018/D-2464 0169). Where a âSt John the Baptistâ was described as having âformer colourationâ [alte Fassung] (401018/D-2464 0151), wood may be assumed but was consistently left out.
In two drawings without details, they are actually lacking for one of them (400405/D-2400 0107 [2]), but the text for the other includes âink brush and lithophoneâ (420513/D-2733 0237). Taking a sample of 52 paintings without material data, nine of them are described with material information that was not extrapolated. At the same time, watercolours (78) and pastels (44) could simply have been copied from âobject typeâ into the field âmaterialâ to allow a normative reading32 â mysterious again
âDimensionsâ list 3,545 entries and 54 objects without measurements. Among them, there are 18 pieces with dimensions in the catalogue text, which were not extrapolated, while 36 objects were indeed catalogued without measurements.33
The following standardised categories are filled in significantly less. After large shares in the field âgenreâ, that is genre (619), history (674), and landscape (899), nearly one sixth has no data here. Since these allocations needed to be assigned by the editors on the basis of the descriptions, data in these fields were reworked in a less than consistent manner, perhaps indicating a loss of interest.34 Not even half of the objects were classified by the editors in the âsubjectâ field, so that 2,125 objects were not given a subject matter.35 Even though there were numerous text data on signature, date, and inscription, the field âinscriptionsâ was eventually ignored altogether and remains empty throughout. These gaps, which only become apparent in a detailed analysis, are therefore not searchable, although the system suggests it.
Transaction data
Object sellers are listed under âseller nameâ (2,065), but no supplementary sources, annotations, or standardisation were given. The listed âseller[s]â refer back to the catalogues, but not to the owner listings which were a mandatory part of each auction catalogue in 1934-1945, but solely to names appearing in catalogue titles, though not consistently.
The title of an auction on 30 March 1939 named the seller Ferdinand von Stumm (1880â1954). He consigned 57 objects out of 177 lots, of which 25 are included in the GPI as part of the auction with an edited total of 74 lots. Five bronzes out of the 25 were annotated by the editors in the field âsellerâ with âBaron v. Stumm, Rauisch-Holzhausenâ. Why and how were these sculptures selected and annotated with information which is actually contaminated, as it is a combination from catalogue title and owner listing?36
A closely curated selection of 28 boxwood sculptures from the large Adolf List (1861â1938) collection were listed with the consignorâs name in accordance with the auction title in 1939.37 In the second part in 1940, a quarter of the 1,214 lots on offer came from the collection âList part IIâ, with two thirds of the 132 lots listed in the GPI, but without the collectorâs name appearing in the âsellerâ field for the 84 objects, referenced only asâvarious ownersâ [Verschiedener Kunstbesitz].
The most extensive data listed under âsellerâ are 214 entries for âBerliner Bankhausâ in an auction at the end of 1940, where Alois Miedl offered paintings on the Berlin market which he had acquired in the occupied Netherlands, using the Deutsche Landvolkbank as consignor. However, the seller information was not included in the following second part of the auction in March 1941 with 104 paintings, among them 76 from the abovementioned âBerliner Bankhausâ.38 To leave out sellers altogether may be the result of an owner listing which included a ministry and a âcompagnieâ, therefore indicating non-private ownership. But that would be an inconsistent interpretation of a source which was not included otherwise. However, the category is inconsistent from the outset, as it does not include information such as âprivate collectionâ, yet counts a company liquidation once under âprivate ownershipâ but not in another case, or fills the field only for part of the saleâs listed lots.39
Overall, there are almost 2,000 missing âseller[s]â compared to 1,676 sales âfrom private ownershipâ, resulting in a neither accurate nor statistically probable analysis, due to catalogue entries being treated inconsistently.
Parallel data
Unlike in the data described above, the editors did analyse additional sources for the following subject groups. While auction houses often produced lists of estimates as printed or duplicated catalogue inserts, result lists sent to subscribers are much rarer. Commercial reasons could speak in favour as well as decidedly against publishing oneâs own results as a marketing initiative for clients.
With regard to the sample of 35 auctions at Hans W. Lange, 33 estimate lists by the firm have been preserved as well as 12 result lists, with German Sales listing digital copies of 27 of the one and three of the other.40 In the GPI, there are thirteen out of 29 auction sales with estimates, but sources are only listed in three instances, all of them referring to the John Paul Getty Museum (JPGM). Were these estimates extrapolated from annotations or lists at the JPGM and was this a different practice from the estimate lists analysed from ten other catalogue copies, and if so, which ones?41 After the process, the editors were able to include estimates for two thirds (62.9%) of lots, but clearly did not look for other catalogues or at least those whose copies are widespread.
External art market periodicals were used for hammer prices and sale prices (these did not publish estimates), mostly Weltkunst (2.128 prices), a publication offering the most comprehensive result lists of the period. In addition, there was the Kunstpreis-Verzeichnis (19) published by the same firm and other periodicals which reported selectively, like Internationale Sammler-Zeitschrift (109) and Pantheon (39). Further prices came from JPGM catalogues (154) and the Berlin museum library (63), therefore from annotations of unidentified auction visitors.
A sale could thus be listed for 2,511 lots out of 3,600 (69.7%); the transfer remained unknown for 1,024 objects (28.4%) and 63 objects (1.7%) were marked as unsold.42
For an auction in October 1940, the editors listed the JBGMâs annotated prices for 73 out of 129 lots; where these were missing, the results published in âPantheonâ (32) and Kunstpreisverzeichnis (14). Nine instances list several prices, indicating a process of comparison and elimination of differences, with matching information not flagged. A second price and source appear to be mentioned only if they differed. The effort is made even though Weltkunst published a comprehensive report, and this publication was the most reliable, since their 2,129 results only required 40 editorial annotations from other publications.43
Cross-referencing seems to have been chosen by the editors, with prices being compared but sources not listed. Users may therefore assume that the price was extrapolated from different sources, even though it is not possible to identify which.
Interpretative data
The final subject group âbuyerâ leaves the secure field of published sources for good. Buyer names were not published for auction sales, with the exception of museum purchases whose publication was typically approved.44 They were recorded, though not publicly, in auction protocols required to be submitted to the police and the chamber of art [Reichskunstkammer], but these are only preserved in rare cases.
The only source can be a catalogue copy with hand-written annotations. Such copies become unique, yet the annotating person is hardly ever identifiable. An analysis involves interpretation of the script as well as the markings, abbreviations, and names, which requires comparison with other unique copies. In our sample however, only annotations from one copy each are listed.
There are only 209 entries for buyers in the sample, that is under 6% of objects. They cover four auctions at Lange, two in October 1940 and two from early 1943. All copies come from the JPGM, though it remains unclear whether the annotations were done by the same person.45 Supplements and comparisons to support the interpretation were not sought and only a small part of the auctions in question has annotations. These can concern a third of lots registered in the GPI (401018/D-2464 and 430127/D-2860), a fifth (401019/D-2463) or just a twentieth (430416/D-2906).
Such amounts set limitations on the possibilities for interpretation. Under âverbatim name of first buyerâ, nine finds are categorised as âillegibleâ,46 59 have an added question mark, and only 83 entries are (supposedly) complete names. The majority are abbreviations and identifiers which are typical for auction annotations in haste.
The standard under âbuyer authorityâ takes these abbreviations as a basis for 166 assumed names, suggested by the Getty Research Instituteâs project group Study of Collecting and Provenance (PSCP).47 Nevertheless, its focus is not helpful. For one, linguistic competence is missing, with absurd letter sequences being accepted, and in addition, there is a lack of scepticism when ânamesâ marked as questionable were still transferred as standard.48
Disregarding the entire context of German Sales, the names âBoernerâ and âBörner [?]â are turned into Berners, though the vowels would seem to be clear pointers to the firm of Leipzig antiquarian dealers.49 Even worse, five variations ranging from âBornh.â To âBernh.â are all read as âBernheim, Georgesâ (1870â1958) as if international trade existed but not the art dealers Walter Bornheim (1888â1971) and the âAryanisedâ firm of Otto Bernheimer (1877â1960).
Both the editors and the study group lacked knowledge in interpreting annotating usage. The abbreviation âge.â or âGeg.â with its variants refers to the opposing party in a bidding battle, not the buyer.50 Professional annotations refer to familiar and frequent buyers by abbreviated names. Here they appear in variations such as âAlbrâ, âBraunâ, âv.M.â, âPirâ, âRehâ, and these are also carelessly transferred to the standardisation. Yet they most likely identify âKommissionĂ€reâ, professional bidders commissioned by third parties: Georg Albrecht (1867â?), Carl Braunstein (1887â?), Harald von MĂŒnchhofen (1899â1981)51, Walter Pirschel (1889â?) and Fritz Rehbein (1890â?).
Abbreviations were also used for well-known dealers. They appear under âAlmâ, âBielâ, âBrozâ, âHinrâ, transferred to standardised versions such as âAlmâ, âBielenburghâ, âBrossettâ and âHeinrichâ52, whereas a more likely reading would be Maria Almas Dietrich (1892â1971), Rudolf Bielenberg (1905â?), Oskar Brozukat (1869â1947), and Johannes Hinrichsen (1884â1971). To discuss these interpretations in detail would go beyond the scope of this article. But standardisation in this category can only be successful with comparisons, otherwise it will stand in the way of finding equal and similar terms.
Conclusion
In 2023, Jeroen Euwe and Kim Oosterlinck analysed price developments in the German art market during wartime based on GPI data. The objective of finding an economic foundation for demand referred back to 11,900 âobservationsâ drawn from csv files, with the analysts being unaware or not in a position to state how many data sets included which prices and how reliable these were.
For the auctioneer Lange, they arrived at an inexplicable 1,841 data sets,53 the third largest sample after the firms of Lempertz (Cologne) and Dorotheum (Vienna) but still excluded from the analysis. The seller âBerliner Bankhausâ however only appears at Lange, where the analysis finds steep rises in prices: âsold at a premiumâ. The analysis missed the next auction from the very same âBerliner Bankhausâ at Lempertz on 5 February 1941/D-2519, where 165 objects were offered with a much more meagre result, so that the highly commendable research is based on individual results which are somewhat uncertain.54
Art market knowledge has significantly expanded since these data were extracted. This becomes abundantly clear even in a comparative sample, because a working copy of the Oppenheim collection sale at Böhler â now also available in digital format â confirms that the reported prices were correct but incomplete, while listing the missing buyers for the remaining lots.
Art market research has since produced more material on prices, actors, authorities, and provenances than the GPI reflects: Its data are from 2011 to 2013.55 The present analysis also demonstrates its inconsistent structure. At the very least they require systematic correction, if not a more desirable large supplementation and reappraisal to make them fit for purpose. For typological searches, the data are indispensable, but by no means infallible.
The data structure which became apparent here, the filled-in fields and the deficits found forecast a multiplication of categorical problems in GPI Remodeled. The complex data model requires structured searches, and a search via indices which are not filled. This is evident where for example a âmaterialâ was not extrapolated so that an object search via category in Remodeled must fail. In this way, users are thrown back to a full-text search and must face the problem of thin data, remaining baffled by a multitude of âno data foundâ.
* I would like to express my gratitude to my colleagues from the AG Digital group for (attempted) searches and for considerations, and the critical readers of my versions, last but not least Dr Susanne Meyer-Abich for her translation.
1 Cf. Getty Research InstituteÂź, CSV exports of the Getty Provenance IndexÂź (June 1, 2022), Sales Contents, Sales Descriptions, Sales Catalogs, https://github.com/thegetty/provenance-index-csv/tree/main/sales_catalogs (120226). First version December 2017, subjected to an âadditional cleaning and normalization work to support linked data transformationâ; currently available version June 2022; concrete differences would need to be investigated.
2 Cf. the categories of the GPI hits via https://www.getty.edu/research/provenance/ on GPI Legacy sub https://piprod.getty.edu/starweb/pi/servlet.starweb?path=pi/pi.web (240226); results do not have URL, reproducible by sales number and four-digit lot number, i.e. âD-2235 0214â for the âRömische Ansichtenâ, which is not an identifier but a full-text search (!), which is why additional results may appear, e.g. D-2400 0107 finds a drawing and a painting F-A442 0107, as paid for with FF 2.400.
3 Per CSV https://github.com/thegetty/provenance-index-csv/tree/main/sales_catalogs#files, sales_contents_8.csv (267 MB), sub title (120226).
4 Bendor Grosvenor, Getty Provenance Index Update (ctd.), Art History News, 02 May 2025, https://www.arthistorynews.com/articles/8709_Getty_Provenance_Index_Update_ctd (160126), explains effects when searching in GPI Remodeled and quotes anonymously a succinct comment: âIt’s fine when you have lots of data but in fact the Getty provenance index is quite ‘thin’ â it has tons and tons and tons of lists of pictures, and (ok there is a bit more, but) that’s about it. So when you create a really fancy structure, all the user really sees is the structure â you can’t really see the small amount of data within, because it’s hidden by all the scaffolding surrounding it.â
5 Cf. the full transcription with field names under GPI Remodeled, Record of sale D-2235 0214, Transcription Identifier 1131909, https://provenance.production.getty.fargeo.com/research/provenance/report/4935ff93-65ef-323f-99a3-a04fdb75a88b (180226).
6 Per CSV https://github.com/thegetty/provenance-index-csv/tree/main/sales_catalogs#files sales_contents_8.csv, sub D-2235,”1938/11/18BLLAN”,”0214″ (180226).
7 Auction Lange 18 November 1938, lot 214, https://digi.ub.uni-heidelberg.de/diglit/lange1938_11_18/0024; on the dinkus Wolfgang Beinert, SchlussstĂŒck, 13 February 2025, https://www.typolexikon.de/schlussstueck/ (170226).
8 Per GS https://digi.ub.uni-heidelberg.de/diglit/lange1938_11_18/0024/ (170226).
9 Weltkunst 27 November 1938, p. 4 and 04 December 1938, p. 4, https://digi.ub.uni-heidelberg.de/diglit/wk1938/0202; no result list from the auction house has been found; list of estimates in the catalogue GS, https://digi.ub.uni-heidelberg.de/diglit/lange1938_11_18/0153 (170226). The final â0024âunder âppgâ, in the Data Dictionaries (apparently version 2017) not described, probably page number of the catalogue PDF in the 2022 version.
10 German Sales â Quellen zu Kunsthandel und Provenienzforschung, arthistoricum.net, UniversitĂ€tsbibliothek Heidelberg, https://digi.ub.uni-heidelberg.de/germansales/ (180226).
11 Cf. in detail the âarchitectsâ Ruth Cuadra/Suzanne Michels, Publishing German Sales, A Look under the Hood of the Getty Provenance Index, 17 April 2013, https://www.getty.edu/news/publishing-german-sales-a-look-under-the-hood-of-the-getty-provenance-index/ (180226).
12 For example, Lynn Rother, DigitalitÀt und Kunstgeschichte/Provenienzforschung, (Symposium) 29 May 2016, Item 5-6, https://digigeist.hypotheses.org/97 (100226).
13 The difference with regard to csv field numbers in the first example is mostly due to the nine field stackings in âmultiple groupsâ for variants; for the second, direct export from the database, for example, four fields on price appear in one (sum/currency/source/date) entitled âsale priceâ; on the relation between field groups no findings to date.
14 Per GPI https://www.getty.edu/research/provenance/, sub Legacy https://piprod.getty.edu/starweb/pi/servlet.starweb?path=pi/pi.web, sub D-2235 0214 (170226).
15 They are spread across âsales_contents_8â und â â_9â (incl. auction Paul Graupe 04 October 1937, the auctioneer being already Lange), display 22,3% of objects, counted by lot; in the GPI split in parts, e.g. five studies by Toulouse-Lautrec as 430127/D-2860 0260[a]-[e].
16 The field âobject typeâ lists 99 watercolours, 2,514 paintings, 61 miniatures, 44 pastels and 273 drawings, but 609 sculptures indiscriminately.
17 Cf. auction Böhler 18 May 1936 (Margarethe Oppenheim collection), https://digi.ub.uni-heidelberg.de/diglit/boehler1936_05_18 (190126); here for example 18 miniatures not included.
18 There are missing (per auction sale date inverted/sales number lot number) e.g. a âDouble coat-of-armsâ (380407 094a) consigned by Albert Prybram von Gladona (1890-1974); bronzes from Adolph List (390328 200 ff.); the sculptures of Helene von Hackelsberger van Eyck (1896-1969) are missing cartouches, reliefs, also a âlight-bearing angelâ [Leuchterengel] (401018 137 ff.); also, the wall consoles (430127 307 f.) from Harry Fuld (1913-1963).
19 The âHead of an Amazonâ, Roman, 1st Century, came from the âGraf Wesdehlenâ collection, sold for RM 850/3.200 to a collector from Hamburg, also prominent; the âHead of Dionysâ was the work pf a copyist, 2nd-3rd Century, did not reach the estimate of 1,500 and remained at RM 1.100.
20 Auction Ball/Graupe 28 November 1930 (Camillo Castiglioni): âDie Sammlung C. Castiglioni, Wien. GemĂ€lde, Skulpturen, Möbel, Keramik, Textilien, Eingel. und beschrieben von Otto von Falke, Berlin 1930â, https://digi.ub.uni-heidelberg.de/diglit/ball1930_11_28_29/; auction Ball/Graupe 28 November 1930 (Camillo Castiglioni; Handkatalog): âDie Sammlung C. Castiglioni, Wien. Handkatalog, Berlin 1930â, https://digi.ub.uni-heidelberg.de/diglit/ball1930_11_28 (150326).
21 Three random samples showed that in just one case Ball 24 April 1931 (Hans Wendland), https://digi.ub.uni-heidelberg.de/diglit/ball_graupe1931_04_24/ and -a the Handkatalog, D-404 and D-405, solely the main catalogue was recorded as sale D-404 with 183 lots, while the Handkatalog D-405 was registered without sale content; but for both the collections of Erich von Goldschmidt-Rothschild (Ball/Graupe 23 February 1931, https://digi.ub.uni-heidelberg.de/diglit/ball_graupe1931_03_23) and Leopold Seligmann (Ball/Graupe 28 April 1930, https://digi.ub.uni-heidelberg.de/diglit/ball_graupe1930_04_28 (150326)) the two editions are included, moreover recorded with different lot totals: for Goldschmidt-Rothschild D-375 with 44 vs. D-376 with 45 and for Seligmann D-85 with 26 vs. D-88 with 18 lots; differences would need further analysis with view to criteria and consequences.
22 Solari, âEmperor Augustusâ also before Cassirer/Helbing 10 May 1928, lot 72 (Oscar Huldschinsky), https://digi.ub.uni-heidelberg.de/diglit/cassirer_helbing1928_05_10/0042/ (120326); four numbers appear in all for this object (D-B5282 0072; D-286 0102; D-18 0102; D-2860 0319); later listed sub li 2696 mue 12279 Mauerbach 608.
23 Miniatures were in part completely added to paintings, in part excluded: cf. auction Lange 22 June 1939, lot 47 f. vs. 49-57; Lange 25 January 1940, lot 334-385; Lange 05 April 1940, lot 131-153; Lange 21 May 1941, lot 131-148; Lange 05 October 1943, lot 331-342.
24 Cf. auction Lange 12 May 1942, lot 75-82 ff. vs. D-2732 0075 ff. intermittent sequence, https://digi.ub.uni-heidelberg.de/diglit/lange1942_05_08/0020/ (120326). KrĂŒgerâs bust portrait âBerliner OberbĂŒrgermeister Geheimer Kriegsrat BĂŒschingâ (lot 80), described as âSchwarze Kreide und DeckweiĂ auf braunem Papier ⊠Lithographiert von Oldermannâ (that is Friedrich Oldermann (1802â1874)) apparently read by the editors as âobject typeâ and excluded as a lithograph, while it was intended to enhance the importance of the drawing as a model.
25 Cf. auction Paul Graupe 25 June 1934, lot 72/D-1341 0072, https://digi.ub.uni-heidelberg.de/diglit/graupe1934_06_25/0031/, vs. auction Lange 04 July 1941, lot 60/D-2604, https://digi.ub.uni-heidelberg.de/diglit/lange1941_07_04/0067/, lot missing in GPI; cf. Sotheby’s, London 07 December 2023, lot 35, https://www.sothebys.com/en/buy/auction/2023/old-master-paintings-day-auction/tronie-of-a-man-wearing-a-gorget (120326).
26 Auction Lange 19 May 1941, https://digi.ub.uni-heidelberg.de/diglit/lange1941_05_19a as âsale catalog D-2569â without âsale contentsâ; see for example Alma Salomonsohn-Solmssen collection per LA ID 297607; Paul Eisner collection et al per LA ID 572593 (li 1951 mue 11772), BLHA, OFP-Projekt, Ein GemĂ€lde-Fund in Israel, 28 March 2023, https://blha.brandenburg.de/blha/de/kontakt/presse/ansicht/~28-03-2023-ein-gemaelde-fund-in-israel; Auguste Pniower/Martha Steinberg collection see Caroline Flick, Warum hat es (bislang) nicht funktioniert? Eine doppelte Geschichte der Suche nach einem entzogenen Bild, in: Retour 09 April 2025, https://retour.hypotheses.org/4373 (230226).
27 Smaller compilations of the subject group under âsecond firmâ (306) to Dorotheum, Wien, or âtitle modifierâ (68) for bundles; the âRömische Ansichtenâ are not included as not recognised as a set.
28 122 names without additional qualification, not repeated after the first instance, cf. e.g. auction Lange 08 February 1939, lots 66-77, https://digi.ub.uni-heidelberg.de/diglit/lange1939_02_07a/0014 (130326) for August Gaul under a header, contains under D-2273 0066 ff. standard data but not life dates, cf. the contrast Vernet as above, for multiplication all data enhanced by the editors.
29 Union List of Artist Names, GRI, Getty Trust, Los Angeles CA, https://www.getty.edu/research/tools/vocabularies/ulan/ (290126), circa 293,000 names and base data for visual artists.
30 Cf. under 198 without Ulan (but 42 âstyle ofâ with Ulan), set throughout as ânewâ in âart_authority_1â as for example Osias Beert [500018195] in 401203/D-2491 0023 or 430416/D-2906 0003; despite life dates in the catalogue for example Otto Rethel [500072448]; for Gauermann [500125953] inserted for 430127/D-2860 0136 f., but not for 420512/D-2732 0041; Blechen [500009469] appears 19 times with Ulan, but not in 420512/D-2860 0014.
31 The following appear: alabaster, watercolour, bistre (2), sheet metal, pencil, bronze, boxwood, Ă©glomisĂ© (1), ivory, enamel, pen, gouache, panel, limestone, chalk, canvas, cardboard, marble, nephrite, oil, paper, pastel, vellum, brush, porcelain, red chalk, sandstone, sepia, stucco, tempera (1), terracotta, clay, ink, watercolour; not searchable per âobject typeâ.
32 Some watercolours with the addition âgouacheâ (18), twice on support (cardboard, paper), once âwatercolourâ; pastels without addition throughout, except one âon canvasâ.
33 These are mostly part lots without details; some centimetre sizes with comma (33) instead of the generally anglicised version with dots as in â14.5 cmâ.
34 In addition, portrait (471), still life (142), animals (122), ten ambiguous entries like for example C.D. Friedrichâs âHĂŒnengrabâ as âlandscape portraitâ (430416/D-2906 0276), probably in error.
35 Given options are allegory (8), architecture (80), interior (84), landscape with figures (515), military (3), mythology (narrative) (28), mythology (figures) (98), religious (narrative) (154), religious (figures) (329), ruins (29), battles (6), seascape (88), studies (2), topographical views (46); some mixed categories and a sole appearance âsportingâ (bowman, bronze 381118/D-2236 0044) suggest editorial control.
36 Cf. Chris Nees, Schloss Rauischholzhausen. Wohnsitz eines Industriellenerben im frĂŒhen deutschen Kaiserreich (Studien zu Schloss und Park Rauischholzhausen 1), Norderstedt 2010. â The auction title named only Stumm in âSchloĂ Holzhausenâ; owner listing has more accurately âRauisch-Holzhausenâ with 57 items, of which five under this seller â once again mysterious, once again unreliable with regard to category.
37 Cf. Sven Pabstmann, Die Biografien hinter den Dingen. Stand und Perspektiven der Provenienzforschung in Sachsen-Anhalt, in: Jan Brademann/Gerrit DeutschlĂ€nder/Matthias Meinhardt (Hrsg.), Sammeln und Zerstreuen. Bedingungen historischer Ăberlieferung in Sachsen-Anhalt (Quellen und Forschungen zur Geschichte Sachsen-Anhalts 21), Magdeburg 2020, pp. 225-282.
38 Cf. Caroline Flick, Verwertungskampagne. Beobachtungen zur niederlĂ€ndischen Kunsthandlung Goudstikker-Miedl (March 2022), https://carolineflick.de/verwertungskampagne/, as well as Sophia Barth-Coorssen, Die Kunst des GeschĂ€ftemachens. Alois Miedl (1903â1970), Eine Biografie zwischen Kommerz und Kunst (BrĂŒche und KontinuitĂ€ten 11), Köln 2025. â Also disregarded was Lange 07 February 1939/D-2271, https://digi.ub.uni-heidelberg.de/diglit/lange1939_02_07 (130326): âDer Verkauf erfolgt im Auftrage eines Berliner Bankhausesâ (Jacquier & Securius) without any entry under âsellerâ.
39 No entry for 390207/D-2271 Berliner Privatsammlung (Max Meirowsky) or 390622/D-2334 Aus einer bekannten Privatsammlung in Berlin Wannsee (Hermine Feist Erben); various for 400618/D-2430 E. Kahlert & Sohn i. L. et al as âfrom a private collectionâ, but 381119/D-2235 Firma Ziffer i. L. et al without entry; for 390208/D-2273 Dr. Bousset et al first under sculptures (!) âfrom a private collectionâ.
40 Bibliographie der Antiquariats-, Auktions- und Kunstkataloge. Bearb. von Gerhard Loh, Leipzig 17 (2010), pp. 97-105 Lange with 29 estimate and 8 result lists; in Astrid BĂ€hr, German Sales 1930-1945. Bibliographie der Auktionskataloge aus Deutschland, Ăsterreich und der Schweiz, FĂŒr die Kunstbibliothek, Staatliche Museen zu Berlin ed. by Joachim Brand and Moritz Wullen, Berlin 2013, http://archiv.ub.uni-heidelberg.de/artdok/2251/1/Baehr_German_Sales_1930_1945_2013.pdf (180226), pp. 163-165 just a âSchĂ€tzlisteâ, eight sheets of a âkopierte Ergebnislisteâ with catalogue copies.
41 Auction Lange 07 April 1938/D-2135, 105 lots; Lange 18 October 1940/D-2463, 129 lots; Lange 12 March 1941/D-2537, 105 lot; the field âestimated price descriptionâ only for bundles (20); according to field names âin German sales, source is name of publication (citation in next field)â, yet the following field âcitationâ here without any entry.
42 âUnknownâ is filled in where no sale result was found, here in error 431005/D-2951 0077a-b Tischbein various; âUnsoldâ offers a price for 30 objects with a JPGM reference which can be confirmed neither by Weltkunst or result lists, except 430416/D-2906-0090a-g Terwesten, Group of seven paintings, RM 10.000/0, unsold, aftersale for RM 4.500 to the Museum in Celle, cf. Susanne Kienlechner, Una versione sconosciuta della tela di Giovanni Battista Tiepolo âNettuno offre doni a Venezia, 2019, http://archiv.ub.uni-heidelberg.de/artdok/volltexte/2019/6572 (030226), pp. 51-54, p. 61 f.
43 Catalogue annotations by Museumsbibliothek Berlin (MBB), probably by a staff member, seem more accurate: In 63 entries there are only four differences to the in-house print version. Auction Lange 25 September 1941/D-2620; https://digi.ub.uni-heidelberg.de/diglit/lange1941_09_25 (020326) with both lists.
44 Here e.g. the buyer of a wooden sculpture Jorhan, Kavalier (380407/D-2125 0095) the museum Liebieghaus Frankfurt M. according to the post-sale report ISZ 29 (1938), p. 91.
45 Even worse, it seems to be impossible to still ascertain which copy was used: Not only is the provenance history of the catalogues â the library copies marked JPGM â not preserved: these seem to indicate only prices and no names with the exception of one copy bis with âprices and many namesâ from the Wilhelm Arntz (1903-1985) collection, collector and expert, cf. 401019/D-2463 http://primo.getty.edu/GRI:GETTY_ALMA21136703200001551; 401018/D-2464 http://primo.getty.edu/GRI:GETTY_ALMA21136720610001551; 430127/D-2860 http://primo.getty.edu/GRI:GETTY_ALMA21129036710001551 with the Arntz copy; 430416/D-2906 http://primo.getty.edu/GRI:GETTY_ALMA21129049620001551 (030326); with gratitude to Mahsa Hatam, Getty Library, 02 March 2026.
46 In standardisation, the number of illegibles rose to 18; hidden in the fragmentary âMus. …brâ (430127/D-2860 0293) is the real buyer: Tiroler Landesmuseum, Innsbruck.
47 Such as the âauthority name for first buyer, established by PSCPâ according to Data Dictionaries, Sales Contents, https://github.com/thegetty/provenance-index-csv/tree/main/sales_catalogs (110225) under the field name âbuy_auth_name_1â.
48 Cf. for example â entry versus âauthority nameâ â 4301027/D-2860 0034 âP.â vs. âP. (London)â; 401018/D-2464 0172 âSchĂŒttln [?]â vs. âSchĂŒttlnâ; 430127/D-2860 0133 âSarbber-Cr. [?]â vs. âSarbber-Cr.â; 430417/D-2906 0070 â[…]mardt [?]â vs. âMarriotâ et al.
49 Six variants from âKuhlâ to âKĂŒbl, Dresd. [?]â transferred to three different standards; including two cases which are probably correct with the buyer Galerie KĂŒhl as âKuhl, Dresdenâ.
50 Cf. for example âGeg. Almâ versus the standardised âAlmâ (430416/D-2906 0136), according to an annotation in private ownership a private collector prevailed against Maria Almas Dietrich.
51 âv.M.â, unmarked, is standardised as âM., vanâ; to be distinguished from âMĂŒnch K [?]â, which in 1940 presumably referred to the firm MĂŒnchner Kunsthandelsgesellschaft â Kameradschaft der KĂŒnstler, the âAryanisedâ business Ludwig Bernheimer KG, Munich.
52 A special case is the note âLangeâ, used ten times and also adopted as standard, apparently used for returns and cancellations (possibly also previous ownership?); a distinction is made from âLangelâ, standardised as âLangel, Albrechtâ, which, according to another annotation, refers to the art dealer Elfriede Langeloh (1882â1960) (401018/D-2464 0155).
53 Jeroen Euwe/Kim Oosterlinck, The German art market during WW II. European Review of Economic History, 10.02.2023, pp. 1-25, https://doi.org/10.1093/ereh/heac024 (030326), p. 9 f. Neither paintings (2,514) nor paintings with prices (2,121) result in the referenced 1,841 sets; though n.b.: Euwe/Ooosterlink (p. 6) use the csv files of the version of 15 December 2017, while the present analysis had to rely on version 01 June 2022.
54 Cf. Flick, Verwertungskampagne âŠ, pp. 150 ff., with regard to a summary of the increases found for Lange by Euwe and Oosterlink of 0,254 to 0,441.
55 Cf. Collaborators, A component of Getty Provenance Index Initiative, German Sales, https://www.getty.edu/projects/getty-provenance-index-initiative/collaborators/ (100326).
Dr. Caroline Flick, historian, works independently on the history of the Berlin auction house Hans W. Lange and the art market during the National Socialist era.
Nur der Text ist unter der Lizenz Creative Commons Namensnennung 4.0 International nutzbar. Alle anderen Elemente (Abbildungen, importierte AnhĂ€nge) sind âAlle Rechte vorbehaltenâ, sofern nicht anders angegeben.
OpenEdition schlÀgt Ihnen vor, diesen Beitrag wie folgt zu zitieren:
Caroline Flick (8. April 2026). Observations with Regard to Provenance Databases II. A Statistics-Driven Look at Data in the Getty Provenance Index Legacy. RETOUR đ . Abgerufen am 17. Mai 2026 von https://doi.org/10.58079/1615f