Ringkloster, en jysk indlandsboplads med Ertebøllekultur


  • Søren H. Andersen


Ringkloster, Ertebølle


Ringkloster. An inland Ertebølle settlement in Jutland
Interim report on the excavations 1969-72

Although the majority of settlement finds of importance to the study of the Danish Ertebølle culture were made in Jutland, for example at Meilgård, Ertebølle, Brabrand and Dyrholmen, our knowledge of the culture in West Denmark is nevertheless very one-sided. This is due to the fact that all Ertebølle sites investigated there so far have been coastal settlements with an economy marked by proximity to the sea: inland settlements proper have not been demonstrated. A possible conclusion would be that the Ertebølle culture had no inland existence in Jutland; but numerous chance finds and a typological analysis of surface collections along the rivers of Jutland (2) have clearly demonstrated that the apparent lack of inland settlements is due not to their actual absence but to the lack of more intensive investigation.

From other parts of the country (Zealand, Lolland, Funen) inland Ertebølle settlements have been known for many years, but these have unfortunately not been published.

As the Jutland settlements have always played a central role in Ertebølle research in Denmark, it is clear that the demonstration and excavation of an inland Ertebølle settlement in Jutland is a matter of particular interest.

A prime prerequisite for obtaining satisfactory results from any inland settlement is circumstances favouring the preservation of organic materials like bone, antler and wood, to permit a valid comparison with the rich material from the coastal settlements. The well-known acidity of the Jutland bogs makes this very difficult to fulfil; with the exception of the Ringkloster settlement described here, no inland settlements with preserved bone and antler artefacts are known in Jutland.

An opportunity to solve these problems arose in 1969, when Forhistorisk Museum was called to the Skanderborg area, fig. 1, where excavation of a drain in a bog near Ringkloster, at the southern end of Lake Skanderborg, had exposed large quantities of Ertebølle artefacts of flint, pottery, bone and antler.

This settlement is not a new one, but was mentioned as early as 1937 by Th. Mathiassen in his publication of the so-called Gudenaa culture (2), and on the basis of surface collection was allocated to that group of Gudenaa settlements which manifested impulses from the Ertebølle culture of the east coast of Jutland (4).

The results of a trial excavation made in 1969 were so promising, that they were followed by a systematic investigation. After 5 years' work it is now possible to give a preliminary account of the excavation so far and its most significant results.

The Lake Skanderborg region was formed at the end of the Last Ice Age by the melting of large glaciers in the Horsens area (5). The terrain is undulating with steep hills, broad valleys and marshy depressions. The soil alternates between moraine deposits and stratified drift. Just around the settlement, the subsoil consists of heavy boulder clay.

In antiquity, Lake Skanderborg was more extensive than it is today. To the north-west, south and south-east, were arms which now appear as peat-filled depressions or marshy river valleys, fig. 2.

The Ringkloster settlement lies at the foot of a WSW-oriented moraine hill on the eastern side of the elongated depression extending from the southern end of the present lake to the south-east, fig. 2, and formerly part of the lake. 600-700 m south of the farm of Ringkloster the bank forms two low promontories, between which the Stone Age settlement was established, fig. 3.

The extent of the settlement is apparent from fig. 3. It comprises an occupation area proper on dry land (now arable) and a refuse layer stretching into the adjacent peat and mud. It is thus not a case of two separate settlements but of two different environments within the same settlement. At the former edge of the lake, the two environments merge. Of the excavated area, 200 sq. m comprise dry land and 105 sq. m bog.

Dry land settlement: The land settlement covers an area of 200 X 75 m along the prehistoric lake, the edge of which lay approximately where the present arable land joins the bog, fig. 3. In this area pits, fireplaces, post holes and possibly also the sites of houses have been demonstrated. There are also large amounts of flint implements, pottery and charcoal, whereas implements of organic materials have not been found.

No special activity areas have so far been demonstrated within the excavated area. Artefacts and other culture relics seem to be evenly spread, but increase in number towards the edge of the prehistoric lake. The type ratios in this area correspond closely to those applying in the bog. As the excavation still covers only a very limited area of this part of the settlement, and as the stratigraphy is complex, a detailed description and analysis of finds must await the definitive publication.

The bog area: In the bog opposite the occupation area is a culture layer, 20-100 cm thick, which can be followed through a zone about 50 m long and 20-30 m wide opposite the central part of the land settlement.

The large quantities of artefacts in this area are most easily accounted for as rubbish from the occupation area on dry land. Besides flint waste and implements, this layer contains large amounts of pottery, implements of bone and antler, thousands of animal bones, antlers, worked-up wood, charcoal, seeds, fish scales, tinder, hazel nuts and acorns.

Fig. 4 is a section illustrating the stratigraphy. The subsoil is sticky yellow boulder clay with stones. Above this is sticky, whitish to light blue calcareous day with thin stripes of fine sand (varved clay): layer 17. This calcareous clay - unusual in a Jutland bog - is responsible for the conditions favouring the preservation of bone and antler. Above it is a 5-20 cm thick grey sand layer, which near the edge of the lake contains a number of small stones: layer 16. Towards the lake this layer rapidly increases in thickness and loses its content of stones. In the peripheral areas it consists of fine white sand with numerous freshwater mussels and snails.

This sand layer continues up onto dry land where it contains numerous artefacts. It also exhibits numerous structures such as fireplaces and pits and must represent the prehistoric surface. The content of culture relicts is every­where large, but the largest concentration is found around the prehistoric lake edge. Towards the lake the content of artefacts falls rapidly and in the actual lake deposits the sand is sterile.

On land, the sand layer is covered by a 30-50 cm thick dark brown to brownish black humus plough layer with a number of artefacts in secondary position. In the bog, it is covered by a 15-20 cm thick layer of brown detritic mud with a little sand, gradually merging into fine-grained, light brown snail mud out towards the lake. Landwards the layer can be followed to the pre­historic lake edge, where it disappears: layer 18. Above, it merges into a dark brown coarser detritic mud with many culture relicts: layer 13-15. Near the bank, the mud contains a number of thin sheets of sand and gravel, which are contemporaneous with the Ertebølle settlement and are probably due to erosion caused by the occupation: layer 14. The culture layer is a clearly defined horizon at the bottom of the detritic mud, fig. 4, and is both above and below clearly limited by layers of sterile peat and mud. In order to show the thickness of the culture layer, all thick-walled, coil-constructed Ertebølle potsherds found within one metre of the section have been projected in, fig. 4. It was nowhere possible to ascertain marked stratigraphical division of the culture layer, which is in principle a homogeneous entity from top to bottom, and contains artefacts of the Ertebølle culture only. Above the detritic mud, layer 13, are alternating layers of brown-red to black-brown humified alder-swamp peat (layer 2-10), which landwards gradually becomes more and more humified and sandy until it finally merges with the plough layer.

In the peat covering the Ertebølle horizon are occasional thin horizons not having the character of true settlement layers, but suggesting brief occupation (hunting visits). The most important of these later visits can be assigned to the Early Neolithic Funnel Beaker B culture. The position of a funnel beaker found in this layer is marked in fig. 4.

The material from Ringkloster is already so large that it places Ringkloster among the largest Danish Ertebølle settlements. It comprises about 26,000 systematically plotted finds, comprising about 850 implements of flint, bone or antler, 3,000 animal bones, 15,000 pieces of flint waste, about 1,500 potsherds and about 6,000 pieces of charcoal, fire-fractured stones and the like.

It is important to note that no significant difference between the land area and the bog area can be demonstrated with respect to the occurrence of individual implement types, cf. the inventory p. 91.

The following survey of artefacts is based exclusively on the rubbish material; this applies also to the illustrations.

Although large quantities of flint waste have been found in all the investigated areas, there is a considerable difference in the amounts found on land and in the refuse deposits. The land concentration is about 500-1,000 pieces per sq. m, whereas the midden concentration is only about 100-150 pieces per sq. m. Implements, on the other hand, are equally frequent, which is probably due to the waste being left where the knapping occurred, while the implements were spread out over the whole activity area, including the midden.

The raw material is Cretaceous flint. The flint technique seems to be generally poorer than that usual for the Ertebølle culture. The waste is characterized by many irregular flakes, and good, regular blades are not common.

Microblades are rare.

Large, regular flakes (raw material for the production of flake axes) do not occur.

The cores are specified in the list on p. 21. Pyramidal or conical forms with a single platform (A cores) are predominant (fig. 6), but blocks with two platforms (B cores) are also represented.

A special type of bipolar core is seen in fig. 7. This form has been used in the production of short thin flakes - presumably for making transverse arrow­heads.

Discus-shaped cores are also present, fig. 8. This type is characteristic of the Ertebølle culture and is presumably (like B cores of the type shown in fig. 7) designed exclusively for making short, thin and broad flakes for transverse arrowheads.

Core-trimming flakes are also common. The various types are listed in the table on p. 26. Commonest are base-platform flakes.

Carinate blades are also present, fig. 9.

The artefact material is specified in the list on p. 91. It is apparent from this that scrapers constitute 12 % of the entire implement assemblage. Simple flake scrapers are common, fig. 10-11, but toothed flake scrapers are also represented, fig. 12. There are also scrapers with a shoulder or rostrum, fig. 13.

The commonest kind of scraper is, however, the simple blade scraper, comprising 25 % of the scraper group, fig. 14-15. If the broken-off ends of such scrapers are included, the type comprises about 50 % of the artefact material. The frequent occurrence of blade scrapers in the rubbish layers makes the type one of the most characteristic forms of the Ertebølle culture at Ringkloster. It is common throughout the rubbish layer, but seems especially prominent in the lower layers.

A few blade scrapers are toothed, or have been fashioned from edge-retouched blades, fig. 15a. One specimen is on a blade with trimming on the back, fig. 15d. One example of a double scraper was found.

Entirely irregular scrapers fashioned from irregular flint pieces without a percussion bulb comprise a special group, fig. 16.

Borers are few in number and comprise 3.3 % of the inventory. The number of borers is fairly constant throughout the refuse horizon, and the individual types are uniformly distributed.

Blade borers without shoulders are the commonest type of borers at Ringkloster, fig. 18, but shouldered blade-borers are also well represented, fig. 17. Oblique borers, fig. 19, and drill borers, fig. 20, are less prominent. Flake borers are also characteristic, fig. 22. Core borers (thick borers), fig. 21, are surprisingly rare (4 specimens) in relation to the representation at other Ertebølle sites.

Burins are very prominent in the Ringkloster material. This group comprises 11.4 % of the artefact material. Medial burins are rare, only one specimen having been found, fig. 23. Angle burins on a transverse facet are not prominent either, whereas angle burins on breaks are very frequent and comprise 50 % of the burin group, fig. 24. The raw material consists almost entirely of good, regular blades, but in a few cases butt ends of symmetrical surface-trimmed flake axes have also been used, fig. 25. Multiple burins without transverse retouch, i.e. specimens with more than one edge, also occur, but are not prominent.

Burins on transverse retouch are also very common. This applies especially to side burins on concave transverse retouch, fig. 26, but more to multiburins on concave retouch, fig. 27, which are the next-commonest burin type at Ringkloster. Less prominent are side burins on straight transverse retouch and multiple burins with and without transverse retouch, which are represented by only a few specimens. Side burins on blades trimmed on the back occur, fig. 28, but are rare. Transverse burins, fig. 29, are also present.

Transverse-retouched specimens are few in number, fig. 30-33, and comprise 6 % of the implement material. Specimens with straight, distal, transverse retouch, fig. 30, are common; the edge can also form a more or less acute angle with the long axis of the raw flake. Specimens with concave transverse retouch, fig. 31-32, are dominant. A characteristic but sparsely represented type is one with concave, distal, transverse retouch and grip retouch, fig. 31. The most frequently occurring form is, however, the one with concave, proximal transverse retouch, fig. 33, which in nearly all cases is on very thin flakes or blades. All this group probably represents raw material for transverse arrowheads. One specimen exhibits double transverse retouch.

Blades with curved distal retouch on the back are uncommon. The rubbish layer has yielded a few trapezes, fig. 34, which can be conceived either as an independent artefact form or as rejected transverse arrowhead rough-outs.

Transverse arrowheads were found in great numbers and are, with 27.4 %, the commonest implement form, fig. 35. The form with straight edge and concave sides dominates, comprising 76 % of the transverse arrowheads. The type with a straight edge and parallel sides comprises 18 % of the transverse arrowheads, while oblique-edged transverse arrowheads comprise only 6 %.

Notched blades and flakes, fig. 36, are relatively uncommon in relation to the other types represented.

Denticulated blades, fig. 37, also contribute to a characterization of the Ringkloster settlement (7.5 %).

This also applies to saws, fig. 38, which are particularly common in the refuse layers. This type is one of the characteristic Ertebølle forms at Ringkloster. Blades with continuous side-edge retouch, fig. 39, side scrapers, and irregular worked flints are also present.

Axes: A total of 92 axes were found at Ringkloster, 76 of which were flake axes (80 % of the axes) and 16 core axes (20 % of the axes). In relation to the investigated area and in comparison with other coeval Ertebølle finds, the number of axes is small. Particularly remarkable is the sparse representation of flake axes with symmetrical surface trimming, which comprise only 26 % of the axe group in relation to the "normal" 40-60 % at coeval sites elsewhere.

Core axes are rare and consist mainly of butt fragments, making typological identification in some cases difficult.

Most of the core axes are atypical, but a few rhombic (oblique) axes and symmetrical axes with bilateral specialized edge-chipping, fig. 40, occur. The last-mentioned form is also known from the upper part of the refuse layer.

Flake axes with trimmed side edges are represented by two types, fig. 41a and b. In the axes from the lowest part of the layer, the percussion bulb has been removed, fig. 41a, while at a slightly higher level another type, fig. 41b, occurs, in which the percussion bulb has been retained at the butt end. The latter type is also longer (8-12 cm) than the former (6-8 cm).

These flake axes are particularly common in the deepest parts of the deposit.

Symmetrical surface-trimmed flake axes, fig. 42, are the predominant form at Ringkloster. This form is especially common in the uppermost parts of the stratum.

Atypical flake axes also occur.

A few 'retouchoirs' have been recovered.

Striker stones of greenstone and spherical crushers of flint, fig. 43, also occur.

Whetstones were also found, being of sandstone, gneiss or schist.

Edge spalls of core and flake axes, fig. 45, are represented; edge spalls of the latter have not previously been described and illustrated. A characteristic type deriving from transverse arrowhead production is bulb spalIs, fig. 46, which have not been demonstrated before in Danish Ertebølle finds.

The Ringkloster material contains large quantities of pottery, as specified in the list on p. 91. The material is so extensive that it affords a good opportunity to analyse the range of vessel forms, and it also exhibits new morphological details and patterns. At least 33 vessels of varying size are represented, 31 being pointed-bottomed pots of Ertebølle type.

Sherds of 'lamps', fig. 53, are rare and occur in only two cases, found in the uppermost part of the stratum.

Most of the pots were constructed using the coil technique, of sand- or granite­tempered clay. The majority of the sherds originated from thick-walled vessels (wall thickness greater than 1 cm). Thin-walled vessels (wall thickness less than 1 cm) are found only in the uppermost parts of the layer, reflecting a tendency towards thinner vessels in the course of the occupation period. The finding of large parts of vessels permits a distinction between different sizes of pointed­bottomed pots, fig. 50-52. Fig. 48-49 give a series of rim profiles showing the variation in the shape of the rim.

Most of the potsherds are undecorated, which is normal in Ertebølle vessels, but a few sherds exhibit a rhombic pattern, fig. 54, stab pattern or double-stab pattern, fig. 56. This ornamentation is unusual in Danish Ertebølle material, and occurs on both thick- and thin-walled pottery. Stratigraphically, these vessels seem to belong mainly to the uppermost and middle parts. Pots with traces of repair also occur.

With respect to vessel construction the material can be classified into typical Ertebølle technique, hammered Ertebølle technique and oblique lamellar technique (62). The thick-walled vessels were mainly made by the hammered technique, whereas the thin-walled sherds most frequently show oblique lamellar technique.

Potsherds in typical Ertebølle technique were found especially in the deepest parts of the stratum, whereas pottery in 'oblique lamellar technique' is known only from the uppermost part. The stratigraphical distribution of the different types of pottery at Ringkloster thus seems to indicate a chronological development (63).

Artefacts made of these raw materials are present in great quantities, yet typologically uniform.

The Ringkloster settlement is characterized by a strong dominance of antler implements and waste, whereas bone implements are rare; cf. list p. 93.

Antler axes are numerous and present in two forms, fig. 58-59: the older has its hafting hole near the burr and is represented by 9 specimens; the younger is hafted through a truncated branch (T-shaped) with 23 specimens. Both types have a straight edge. In the refuse horizon they occupy different levels, showing that a development has taken place, fig. 57. In no case were the two forms found at the same level in the same stratum.

From the diagram fig. 57 it is further apparent that antler axes were absent from the latest phase of the Ertebølle settlement at Ringkloster. This feature of the western Danish Ertebølle culture has not been observed before, and similar circumstances have so far not been demonstrated at other localities. In comparison with other coeval Ertebølle finds the number of antler axes is very large. In several cases the wooden handle is preserved, fig. 59.

Chisels or burnishers, fig. 60b-c, have not previously been described in Danish Ertebølle finds. This form is made of sawn-off tines of red deer antler at the tip of which a long, narrow, tongue-shaped edge has been fashioned on the concave inner surface.

Strikers, fig. 61, are very common.

A number of shoulder blades from which discs have been cut were found, fig. 62. This type is especially prominent in the deepest and middle levels of the deposit. Shoulder-blades of aurochs, red deer and wild boar were utilized. A single bone disc, fig. 63, was actually found.

The Ringkloster find contains only one bone point, fig. 64, which is otherwise one of the most frequently occurring artefacts at Ertebølle coastal-sites. Waste products from the manufacture of bone points are also very rare, fig. 65.

Bone knives, fig. 66, are also a new type, not previously demonstrated at Ertebølle settlements in Denmark. The raw material is aurochs and elk rib.

One smoother, fig. 67, is present.

6 tusk knives on the lower canine teeth of wild boar were found. Other antler and bone artefacts were fragments of smoothed red deer antler picks, fig. 60a.

Apart from the implements of bone and antler the bog contains large quantities of antler waste, in particular from axe-making, cf. fig. 68.

The refuse horizon contains a large quantity of sharpened and severed hazel branches, between 15 and 200 cm long.

15 antler axe handles were also found. Natural, three year old branches of hazel (most commonly), alder and rowan, with the bark intact, were used. The preserved remains are 4-35 cm long, fig. 57. None of the hafts exhibits special fashioning. The material also comprised a large fragment of a bow, fig. 69, of slowly grown elm.

Other finds were fragments of oar blades, axe and arrow shafts and club-like pieces of wood of unknown use.

The fact that several types either change gradually or are replaced by others in course of the occupation period shows that the Ringkloster settlement was used over a long period. The oldest part of the culture layer at Ringkloster can be equated typologically with Norslund layer 2 (90). This phase has been C14 dated (Haldrup Strand K-1612) to 3680 ± 120 B.C. The middle part of the deposit is analogous to Norslund layer 1 and Flynderhage, which is C14 dated to 3280 ± 100 B.C. (K-1450).

The most recent part of the culture layer (the horizon without antler axes) cannot be dated more accurately, since comparative material from other Ertebølle sites is at present lacking. On stratigraphical evidence, it can merely be said to be younger than the middle horizon, which is dated to about 3300 B.C. In connection with the excavation of the refuse horizon, C14 determinations have been made for charcoal from different levels. The deepest specimen, taken near the base of the stratum in direct contact with an antler axe with hafting hole near the burr, yielded a date of 3660 ± 110 B.C. (K-1652). Another specimen was taken just above the bottom (K-1765) and gave 3550 ± 100 B.C. A specimen in contact with a T-shaped antler axe and taken from the middle level (K-1653) yielded 3540 ± 100 B.C. Finally, charcoal from the upper levels (K-1654), yielded a date of 3370 ± 100 B.C. There thus seems to be close agreement between the stratigraphical, typological and physical determinations.

Animal bones from the 1969-72 excavation campaigns - 1,685 specimens - have been identified; cf. list p. 94. Animal bones occurred everywhere along the edge of the lake, but it is characteristic that they were very frequently found in discrete concentrations, presumably representing rubbish from the individual periods, fig. 71.

Gathering is documented by the presence of thousands of nut shells.

The bog also yielded seeds and fruits of dogwood, hawthorn, yellow flag, buckbean, yellow and white waterlily, pondweed, bittersweet, burr reed, lime and oak. Apart from lime, hawthorn and oak, all the seeds and fruits are from bog or water plants. It is not evident whether their presence is due to gathering, growth on the spot or both.

The material reflects a surprisingly one-sided economy, almost exclusively based on hunting of wild boar and pine marten (cf. fig. 70 and the list on p.94), while fishing and fowling have not played any significant role.

Only one previous instance of a correspondingly one-sided economy has been demonstrated in the Danish Ertebølle culture, at Hjerk Nor (110).

Large game (aurochs, elk, red deer, wild boar and roe) comprise 86 % of the bone material. Fur-bearing animals (marten, polecat, badger, otter, wolf, fox, wild cat and beaver) comprise 11 %; birds (swan and white-tailed eagle) 0.6 % and fish (pike, bream and perch) 1.6 %. The only domestic animal is the dog.

lf the bone material is related to the locality (freshwater lake surrounded by forest), the sparse occurrence of fish and fowl is very striking.

Within the group of large game the wild boar is unusually prominent with 74 %; red deer comprises 16 %, aurochs 7 %, roe 2 % and elk 0.6 %. The boars are large, strong animals of all ages. Many animals are over 5-7 years, but sucking pigs (5-6 months) are also represented. Bones of aurochs derive from both very large bulls and smaller individuals (females). A number of bones are so small that they can easily stem from domestic cattle.

The red deer is not prominent. The number of bones and unshed antlers is small, and this species seems to be underrepresented in the bog.

Roe deer are surprisingly rare - especially as they must have occupied the same biotope as red deer.

Among fur-bearing animals there is an overriding predominance of pine marten (77 %), but otter is also well represented (14 %). The other fur-bearing animals are represented by only a few bones. The strong dominance of pine marten is characteristic of the Ringkloster settlement and has not previously been noted in a Danish mesolithic find. The bones of these animals are often found in discrete piles in the midden. As entire skeletons seem to be present in several instances, it looks as if the animals have been thrown into the 'lake' after being skinned. On most of the marten skulls there are distinct cross-cuts just above the frontal ridges, fig. 72. The low occurrence of beaver (1.6 %) is striking, for large amounts of beaver-gnawed stricks and trunks show that this species was common in the prehistoric lake when the settlement was occupied.

With respect to the season of settlement, the high proportion of unshed antlers in relation to shed antlers shows that occupation was in winter (October-April). This is consistent with the intensive hunting of fur-bearing animals, whose skin is best in winter.

The copious presence of hazelnuts points in the same direction (September­October). An analysis of age distribution of wild boar within the two first years of life show that they were all killed in the period September to May.

The present evidence indicates that Ringkloster was a seasonal, winter settlement, and that the economy was specialized for hunting wild boar and fur­bearing animals.

Great interest attaches to the demonstration of a rib of bottle-nosed dolphin. This marine provides proof of contact between Ringkloster and the coast. As the Ringkloster settlement is situated equally far from Brabrand Fjord, Norsminde Fjord and Horsens Fjord, fig. 1, it is not possible to be more precise. The occurence of stab-decorated pottery at Ertebølle settlements at Norsminde Fjord suggests that the Ringkloster settlement's area of contact should be sought there. Neither is it possible to know whether the rib is a manifestation of direct contact, i.e. whether it was brought from the coast to Ringkloster by one of the inhabitants during one of the seasonal moves, or whether it arrived by other means, for example in trade. It seems most likely that the inhabitants brought a piece of whale blubber from the coast in connection with a seasonal move - either as provision or as fuel for the blubber lamps.

With respect to the position occupied by Ringkloster in a settlement pattern, two possibilities present themselves. The settlement has either served as a seasonal settlement for a single group of people practising seasonal migration between lake and coast, or as a 'common' area of exploitation for several of the major east Jutland coastal settlements.

In conclusion, it may be stated that Ringkloster is the first inland Ertebølle locality in Jutland with preserved organic remains, and the artefact assemblage exhibits a number of special features distinguishing it clearly from the coeval Ertebølle coastal settlements. The settlement was seasonal and winter, with a strongly specialized economy directed to the hunting of fur animals and wild boar. It is presumably against this special background that the characteristic features of the artefact assemblage should be seen. A systematic comparison with a coeval settlement must, however, await completion of the Ringkloster excavation.

Søren H. Andersen
Aarhus Universitet, Moesgård







Andersen, S. H. (1973). Ringkloster, en jysk indlandsboplads med Ertebøllekultur. Kuml, 23(23), 11–108. Hentet fra https://tidsskrift.dk/kuml/article/view/97017