Figure 1. Internal analysis of tool markings on formerly soft wall deposit in Nung-kol Cave, South Australia. Where striation patterns are preserved, five apparently different tools have been identified, numbered 1-5. All marks were executed from top to bottom, with limestone clasts. The cross-sections of the tips (perpendicular to direction of movement) on two of the tools were determined, and are shown below (tool 1 and tool 2).


Figure 2. Principles of the relationship of total production of an archaeological phenomenon s-alpha to its surviving instances s-beta as a function of angle phi. These principles are the basis of taphonomic logic.

The obvious qualification of ethnographically and iconographically derived interpretations is, as mentioned above, that such mythologies must not be created in the name of science; it is not the business of science to create myths about the past (Bednarik 1992c). Science certainly does have a role to play in palaeoart studies, and although the application of scientific practice in rock art studies is a recent development, good progress has been made in recent years. Indeed, this discipline has progressed much more in the last fifteen years than in the previous two hundred years. An example is the recent proliferation in serious dating work. Among the specialised subjects being developed now are various technological studies and physico-chemical analyses; the identification of various types of residues (for instance, binders, proteins, lipids, extenders etc. in rock paints, or organic inclusions in mineral accretions over petroglyphs) (Clottes et al. 1990; Cole and Watchman 1992; Ridges et al. 2000); the use of computerised programs for image analysis, manipulation and colour standardisation (Rip 1989; Bednarik and Seshadri 1995); nano-stratigraphy of accretionary deposits and paint residues (Bednarik 1979; Watchman 1992); microscopic study and ‘internal analysis’ of tool marks in rock art and portable art (Marshack 1985); discrimination of anthropic and non-anthropic marks on rock and portable objects, and relevant ethology (Bednarik 1991d, 1993b); replication studies (d’Errico 1991; Bednarik 1997); erosion and microerosion studies (Bednarik 1992d); relationships between phylogenic and ontogenic development of logic and symbolism; the psychology of iconicity and its decipherment; concepts of type and typicalness in pre-Historic art, symbolism and psychology; distinction between mental and artistic representations; the application of taphonomic logic at both technical and epistemic levels (Bednarik 1994a); epistemology in the formulation of theories and in the interpretation of palaeoart; valid applications of statistics in the discipline; sound utilisation of universals in palaeoart studies (Bednarik 1990/91); and other experimental approaches. A few such approaches that can significantly assist interpretation are discussed below.



This term was coined by Alexander Marshack to describe the microscopic study designed to determine various technological details of engravings, such as direction of tool application, handedness of operator and multiple tool application. Marshack’s principal motivation was to find a way of recognising notational markings (Marshack 1972, 1985, 1989), but while the methods developed from his pioneering work have provided superb insights into questions of how engravings were made, they cannot furnish conclusive evidence for or against interpretation as notation (d’Errico 1989; Bednarik 1991e).

Apart from this one proviso, internal analysis has been very successfully used and tested by both d’Errico (1988, 1991, 1994) and myself (Bednarik 1984, 1988, 1992e). Whereas Marshack and d’Errico applied this form of analysis only to portable objects, I extended it to rock art, especially engravings and finger flutings in limestone caves.

The principal difference between the work of Marshack and d’Errico is that the former researcher conducted little if any replicative work to underpin his theoretical constructs and pronouncements, relying essentially on visual judgment acquired from close familiarity with the materials. D’Errico has undertaken a great deal of experimental work in mark production and related fields. He also made extensive use of the scanning electron microscope, whereas Marshack and I only had access to optical microscopy. To examine striations and other details, d’Errico used not only the reflected-light microscope, he also created transparent cast replicas to study with a transmitted-light microscope, and resin casts to examine by scanning electron microscopy. He focused on incidental striae where tool aspects other than the point impacted upon a surface, and on changes in the pattern of contact, e.g. in response to variations in curvature of the engraved surface, or at points where the burin splintered.

D’Errico has shown through painstaking replicative studies that a good understanding of such minute tell-tale features in engraved lines, which can survive for many millennia on some surfaces, provides much reliable and testable information about the circumstances of mark production. The repeated application of the same tool point can be identified with great confidence, and it is often possible to determine the direction of tool application. Precedence in superimposition is another easily resolved issue, on sufficiently well preserved specimens.

It is often also possible to determine the type of material that was used in producing an engraved or incised (i.e. abraded) marking on rock, particularly on relatively soft rock (e.g. in a limestone cave, or on low metamorphism products such as slate, phyllite or schist). Applying blind tests I showed that an experienced observer can identify the experimental marks of metal points of utilitarian objects in nearly all cases, provided that the points had not been broken. The jagged ends of broken steel, for instance, yield marks resembling those of some stone points. Therefore it cannot be determined with confidence whether an engraved mark was made with a stone point, even if that does appear to be the case, whereas it is certainly possible to recognise the marks made by unbroken metal points with great confidence. Using this method I was able to establish that some of the supposedly Palaeolithic engravings in the Côa valley of northern Portugal had been made with steel points, and could therefore only be of final Holocene antiquity (Bednarik 1995a).

In some cases the method does not even involve a need for magnification (Figure 1, above). For instance, finger flutings in limestone caves are sometimes so well preserved that the direction of movement of the fingers is ascertainable from tiny tear-marks and other indicators (Bednarik 1986). Similarly, rock paintings would yield information about their execution by this approach, particularly where the paint was applied with fingers.

Marshack’s internal analysis has become one of the most effective tools in the scientific interpretation of mark production processes, but I contend that the primary purpose for which it was initially conceived, the recognition of notational markings, is not served by it. While it is certainly possible to determine that two or more marks were made by the same tool point, it is impossible to know if two different tools were used in any two marks. This is because stone tool points are generally of non-symmetrical and irregular shape, so if the tool is put down briefly between applications, or if the grip or direction of application, even the angle of application, is changed between two marks, it may be impossible to recognise the repeated use of the same tool. Since a principal characteristic of notations is that a sequence of similar marks was made with interruptions, often using different tools, it follows that the issue cannot be resolved conclusively by this method. It cannot be demonstrated that two morphologically different markings could not have been made by the very same tool point.



One of the most serious limitations of statistical analyses in rock art research is that posed by the inherent subjectivity of the primary data. Irrespective of the actual method used, statistics that address the content of rock arts always involve a taxonomy of motif elements, because the grouping of motifs perceived to be similar is a prerequisite for such treatment. Yet any such ‘taxonomising’ process is entirely based on the iconographic perceptions or graphic and depictive conventions defining the researcher’s own system of reality. It does not reflect the artists’ graphic cognition.

Empiricists seem to realise part of this problem, but their belief in their own objectivity still allows them to perceive meaningfulness in the arbitrary selection of criteria to create taxonomies. It would initially appear that nonfigurative art, consisting of ‘geometric’ arrangements, is more accessible to statistical analysis than figurative art. In fact the opposite is true, because while there might be some merit in assuming that the observer shares iconic perceptions of the artist, that cannot apply in noniconic art. No two rock art motifs are identical in every respect, yet in order to prepare a site’s assemblage for statistical treatment, apparently similar motifs must be grouped and considered together. Without this process no statistical treatment of any corpus of rock art is possible. Each and every motif possesses thousands of variables, and the analyst must decide which ones are to be considered crucial in determining its taxon. Otherwise there will inevitably be as many motif types as there are motifs at any one site. Among the characteristics available for selection are metrical indices, qualitative or formal aspects, aspects of the motif’s relationship to the site, to other motifs at the site, to the motifs at other sites, to the topography of the site, to petrological or past vegetational or hydrographic aspects of the site or of the surrounding landscape, spatial or syntactic context, the identity or status of the artist (is a circle engraved by a man the same motif type as one engraved by a woman?). Not only can single elements have wide ranges of ‘meaning’, definable only in terms of pre-Historic context (consider Munn 1973), the number of characteristics of each motif is practically unlimited. Many of those characteristics that defy archaeological definition may be crucial in identifying motif types (e.g. the sex of the artist). Thus the parameters the analyst chooses will inevitably reflect her/his personal, cultural, historical, ethnocentric and cognitive biases. Consequently the information so derived is useful only in studying the analyst’s own culture and cognition, and in studying the way in which s/he applies these in examining the surviving graphic traces of cognitive systems to which s/he has no cognitive access. However, that information is not scientifically relevant to the study of the pre-Historic culture concerned, except for purely descriptive purposes. It cannot be expected to provide valid data for statistical analysis of the palaeoart.

This is one of the two fundamental objections to statistical interpretations of rock art, the second involves the application of taphonomic logic (see below). Both these concerns are fatal for the archaeological belief that rock art interpretation is possible through sophisticated statistical treatment. It needs to be emphasised that there certainly is a place for such treatment, namely in the empirical description of what there is on the rock surface today, and how a conditioned alien researcher perceives this legacy. For purely descriptive purposes such a site inventory is of limited usefulness (e.g. in conservation work), but it is not an adequate basis for further analysis. It must not be regarded as the raw data for any kind of ‘scientific’ treatment. For such purposes it is far too crude and inadequate. Any form of analysis of a corpus of rock art would have to be based on the whole of the art that was produced, and it would have to break down this whole into temporal units. If that is not done, or not possible, then statistical data about rock art are not scientific data.

Archaeologists have argued that rock art motifs are artefacts and should therefore be treated just like any other artefacts, and the perceived rigour of statistics should somehow conjure up valid interpretations. But no archaeologist would ever add up all the lithics from an excavation, lumping together those from all levels in order to compare the total with similarly derived totals from other excavations. It is clear to all that such a procedure would serve no useful purpose in the case of excavated artefacts, it would be the antithesis of all archaeology stands for. Yet in rock art studies this is precisely what many archaeologists have proposed and done. It is therefore essential to emphasise that such an exercise is for all realistic purposes entirely worthless. Worse still, it is severely misleading.



Metamorphology, the scientific version of archaeology, is a logic-based, refutable system of reviewing archaeological information to determine whether archaeological propositions could have scientific legitimacy. It is developed especially from taphonomic logic, which in turn hinges on the concept of cumulative data loss as a function of time (the principle is depicted graphically in Bednarik 1994: Fig. 2). It replaces uniformitarianism as a unified theory of archaeology (cf. Cameron 1993).

One of the greatest heuristic stumbling blocks of orthodox archaeology is that it tends to treat evidence — or what it calls the ‘archaeological record’, an essentially meaningless concept (Bednarik 1994c) — as a kind of random sample, as if it amounted to a representative selection of variables defining a particular culture. In the case of rock art, this is rendered particularly incongruous by two factors: that major rock art sites are usually cumulative records in mostly two-dimensional space (although there are exceptions, such as nano-stratigraphic sequences), and that the scientific dating of their chronological components continues to remain extremely difficult (Bednarik 1990-91). In other words, the rock art at such sites belongs to different periods, the artists of which contributed to the same corpus, perhaps reacting to pre-existing art at the site. Traditional archaeological approaches are practically pointless here, and the determination of what actually constitutes a valid sample is often extremely difficult, if indeed possible.

Expressed in epistemological language, there is a dependency relation called a supervenience: one set of properties, forming the historical event, is supervenient on a second, represented in the selected sample. There could not be a difference in the first without there being a difference in the second, though there could be a difference in the second with no difference in the first. Thus the relationship of the two sets of properties has to be explored by alternative means, not by direct deductive reasoning.

Metamorphology is the science of how forms of evidence change with time to become the forms as which they are perceived or understood by the individual archaeologist today (Bednarik 1995b). In accounting for the considerable gap that exists between the reality of what actually happened at some point of time in the distant past, and the abstraction of it as it is perceived by an archaeologist, metamorphology has to take into consideration such a myriad of factors that it cannot be expected to provide precise interpretations. Like refutation in general, it instead provides us with models of what is unlikely to be valid, and so strengthens archaeology by weakening its dogmas. It also eschews the concepts of an ‘archaeological record’ and of a collective knowledge of the discipline.

The most obvious of these factors accounting for metamorphology, and thus for the gap between archaeologists’ constructs and what really happened in the past, is taphonomy (Efremov 1940). Taphonomy deals with the logical underpinning of the idea that the quantified characteristics of a record of past events or systems are not an accurate reflection of what would have been a record of the live system or observed event. Taphonomy distorts archaeological evidence systematically, and it does so in forms that have often not been appreciated by archaeologists. Indeed, after the palaeontological concept of taphonomy was introduced into archaeology a couple of decades ago, it was soon misunderstood and effectively became a practice resembling actuopalaeontology — which ironically taphonomy was originally intended to replace (for an excellent discussion of this point, see Solomon 1990). Hence the potential of taphonomy in archaeology has remained significantly under-utilised.

In rock art science, taphonomy is the study of the processes affecting rock art after it has been executed, determining its present appearance and statistical properties. Taphonomic logic is a form of logic viewing rock art as the surviving remnant of a cumulative population that has been subjected to continuous degradation which selects in favour of specific properties facilitating longevity (Figure 2, above). It inevitably leads to the concept of the taphonomic threshold, which is the point in time at which instances of a specific class of material remains either begin to appear, or begin to appear in significant numbers. Before their respective taphonomic thresholds, all classes of material remains experienced a taphonomic lag time. This is the period during which the phenomena resulting in the material remains in question did exist, but from which we have found no such evidence — or so little that it is usually explained away as a ‘running ahead of time’ (Vishnyatsky 1994), or as incorrectly dated or identified. Until examined closely this may seem a minor issue, but for most classes of material remains, the taphonomic lag times are well in excess of 90% of the phenomenon’s historical duration. Indeed, taphonomic lags in excess of 98% and 99% are quite common, they apply for instance to seafaring evidence (Bednarik 1999), and to most organic remains used as artefacts (e.g. hide, bark, wood, fibre, resin). Taphonomic logic is quantifiable, at least to the extent that it can be expressed as an integral function (Bednarik 1994a: 73).

In rock art, taphonomic lag times remain largely undetermined, but they would certainly differ greatly according to the climate, type of rock art and type of support rock. The taphonomic threshold of beeswax rock art in the Northern Territory of Australia (Nelson et al. 2000) has been estimated to be about 800 years BP (Bednarik 2002). One of the most common forms of surviving rock art consists of red paintings in sandstone shelters, and their threshold is believed to be in the order of 8000 years. For most other rock paintings the taphonomic threshold is lower. However, there are notable exceptions, especially art sites located in deep limestone caves offering extraordinary preservation conditions. In other words, the Palaeolithic ‘cave art’ traditions of Europe are known to us only because some of the art of the societies concerned was placed in ‘fluke preservation conditions’. That evidence would probably not have survived elsewhere.

The ability of petroglyphs to survive in the open, i.e. exposed to precipitation, is governed largely by the rock they were made on. Those on limestone have a taphonomic threshold of well under 2000 years (Mandl 1996), while those on granite can easily survive from the Pleistocene, and recent dating evidence suggests that their threshold might be in the order of 30 000 or 50 000 years under some conditions (Bednarik 2001). Other relevant variables are climate and geochemistry. Taphonomic logic demands that rock art of the various types was produced before all of these thresholds, but evidence of it should be either unavailable or extremely rare. Occurrence and distribution, as well as various characteristics of surviving rock art are all determined by taphonomic factors. Therefore it would be meaningless to state that a particular tradition produced only deep line petroglyphs, or painted only in caves, or left no open-air engravings. All characteristics of rock art that might contribute to their longevity (e.g. depth, location, type of rock support, morphology of site, composition of paint) are of no relevance to defining a tradition, because taphonomy selects in favour of them. For instance, the probably most common technological form of petroglyph is the sgraffito, made by the removal of a patina or weathering zone to reveal a differently coloured surface beneath. Sgraffiti tend to be obliterated by repatination processes within two or three millennia, therefore it is pointless to observe that the earliest petroglyphs of a region are consistently those that are deeply engraved. This observation, while valid, leads to misinterpretation of the sequence, unless moderated by taphonomic logic.

The mechanics of taphonomic logic are rather more complex than indicated in the present brief comments, but it must be emphasised that they are of crucial importance to the interpretation of primary data about rock art. This is the most important methodological tool so far developed in the interpretation of rock art data, and indeed, in archaeological interpretation of data and in hypothesis building. This form of logic is a recent introduction, it remains misunderstood by most archaeologists, yet it is crucial in transforming archaeology into a science. In rock art research, its introduction is probably easier, because the number of materials to be understood is quite limited, and also because the discipline is not as yet so much driven by, and committed to, dogma.

R. G. Bednarik, 2001