Hidden in plain sight: A microanalytical study of a Middle Stone Age ochre piece trapped inside a micromorphological block sample

A complete Middle Stone Age ochre piece was unintentionally collected and fully preserved within a micromorphological block sample intended to characterise a 74 ± 3 ka occupation horizon at Blombos Cave, South Africa. Previously recovered ochre pieces from the same stratigraphic context (Still Bay) have displayed intricate modification patterns with significant behavioural implications. Yet, in the case of the trapped ochre, a direct visual assessment of its surfaces was impossible due to its impregnated state. In this study, we demonstrate how we successfully reconstructed three‐dimensionally and characterised the block‐sampled ochre piece using high‐resolution microcomputed tomography scanning coupled with a range of microanalytical techniques, including optical petrography, micro‐Fourier transform infra‐red spectroscopy, micro‐X‐ray fluorescence and micro‐Raman spectroscopy. Through a morphometric analysis of the score marks present on the trapped ochre's reconstructed surface, we were able to assess the types of modifications and the nature of the actions that created them. Our results show that a block sample‐based study of archaeological artefacts allows for a comprehensive assessment of their depositional and microcontextual setting, their external morphology and microtopography, as well as their internal texture and geochemical properties. We suggest that this type of context‐sensitive, multiscalar and multidisciplinary investigation may also prove beneficial in the study of conventionally recovered archaeological artefacts.


| INTRODUCTION
During the last decade, archaeological micromorphology has been successfully applied to an increasing number of Middle Stone Age (MSA) sites in southern Africa (e.g., Goldberg et al., 2009;Haaland et al., 2017;Karkanas et al., 2015;Larbey et al., 2019;Marean et al., 2000;Miller et al., 2013Miller et al., , 2016. Due to excellent preservation conditions, sediment samples collected at many of these sites typically contain large amounts of occupation debris derived from past human activities. Occasionally, larger artefacts (>1 cm) are also incorporated into the samples. Most often, this occurs as a result of an intentional sampling strategy in which archaeological deposits filled with occupation debris (e.g., fragments of bone, stone and organic material) and campsite features (e.g., hearths, shell middens and bedding) are purposefully targeted for microstratigraphic analysis.
On rare occasions, however, complete artefacts may be unintentionally sampled. In our case, an intact object (ca. 60-mm long) was collected as a part of a micromorphological block sample (BBC-13-16) originally intended to characterise a 74 ± 3 ka occupation horizon within the MSA sequence at Blombos Cave (BBC), South Africa (Figures 1 and 2). After this block sample was impregnated and cut open with a rock saw, we observed that besides its calcareous and quartz-rich sediments, it also contained a bright red, ironrich rock fragment (Figure 3a). The profile shape and striking red colour of the fragment were clearly visible in the cut block sample Morphologically, the trapped ochre piece resembles previously reported modified ochre artefacts recovered from BBC ( Figure S1). In the thin section, the profile of the trapped ochre shows a highly irregular surface topography (Figure 3d). Its stratigraphic and depositional setting links it to the Still Bay technocomplex, an MSA phase known in BBC for its particularly rich ochre assemblage, including two pieces engraved with cross-hatched designs (Figures S1 and S2) (Henshilwood et al., 2002).
On the basis of these initial observations, we found it likely that a modified MSA ochre piece had been inadvertently collected and incorporated inside the BBC-13-16 micromorphological block sample.
However, any direct visual inspection of its overall morphology and surface topography was impossible due to its impregnated, blocksampled state. To solve this problem, we decided to scan the entire micromorphological block, including the ochre piece within it, using X-ray microcomputed tomography (micro-CT). A micro-CT scanner generates multiple X-rays of physical objects that can be reconstructed into cross-sections or virtual three-dimensional (3D) models in a nondestructive way. In micromorphology, it has previously been used to document and inspect the internal composition and structure of a wide range of materials, including oriented soil and sediment samples (Ngan-Tillard andHuisman, 2017, Villagran et al., 2019).
This study aims to showcase how archaeological artefacts, when trapped inside impregnated block samples, can be digitally reconstructed in 3D. Furthermore, the incidental collection of the trapped ochre piece also presents us with a unique analytical F I G U R E 1 Site map of Blombos Cave. The red line shows the location of the sampled section wall (see Figure 2). Yellow arrow: Location of block sample that contained a red ochre fragment [Color figure can be viewed at wileyonlinelibrary.com] opportunity to study these important MSA artefacts from a range of new perspectives. For this reason, we use here a combination of archaeological micromorphology, 3D field photogrammetry and highresolution micro-CT scanning, combined with a suite of microanalytical techniques (micro-Fourier transform infrared spectroscopy [FTIR], micro-X-ray fluorescence [XRF] and micro-Raman), to characterise the trapped ochre, both in terms of its depositional setting, its morphology, its microstructure and geochemical composition, as well as the nature and location of anthropogenic surface modifications. Ochre is a colloquial term used to describe a series of ferruginous rocks that were intentionally collected and modified by humans from ca. 300 ka (Barham, 2002;Hovers et al., 2003;McBrearty & Tryon, 2006;Watts et al., 2016;Watts, 2010). These rocks, clays and sediments typically contain varying amounts of iron oxides (generally FeO,FeO 2 or Fe 2 O 3 ) or oxide−hydroxides (α-FeOOH) that can produce yellow, orange, brown, red, purple and black-coloured streaks (Barham, 2002;Cornell & Schwertmann, 2003;Erlandson et al., 1999; F I G U R E 3 Documentation of the micromorphological block sample BBC-13-16. (a) Image scans of the four slabs that were cut from the original hardened block sample. The red ochre fragment is located at the top left corner. The yellow box indicates the part where a thin section chip was cut.  Hovers et al., 2003;Watts, 2002Watts, , 2009. In southern Africa, its occurrence has been reported from ca. 160 ka (Marean et al., 2007) and from where its presence has then become ubiquitous at sites across this region (Watts, 1998(Watts, , 2009Wolf et al., 2018).
The antiquity and abundance of ochre at MSA sites have often been associated with the emergence of symbolic behaviours (d 'Errico et al., 2003;Henshilwood, 2007;Watts, 2009). Yet, the precise nature of symbolic pigment is difficult to assess from the current archaeological record, for example, as paint on human skin or artefacts. However, recent research on the operational chains of ochre collection and processing has provided insight into the behavioural and technological aspects behind pigment production. Studies of modified MSA ochres typically show striated surfaces from intensive grinding, which are widely assumed to represent evidence for F I G U R E 4 Top: Photos and three-dimensional (3D) models (yellow) showing the front and backside of the modified ochre piece SAM-AA 8937. Yellow rectangles indicating areas selected for close-up photos and corresponding 3D microtopographic models (seen in 2D and 3D view) of: (a) Deep and wide scoring, (b) scoring, (c) scoring (line 1) and striations (line 2 and 3), (d) striations and (e) microstriations. The location and direction of crosscuts and from where profile heights, lengths and angles were measured, are indicated with red lines. Multiple measurements were conducted along with the same scoring or striation (indicated by line numbers) [Color figure can be viewed at wileyonlinelibrary.com] pigment (powder) extraction (Hodgskiss, 2010(Hodgskiss, , 2012. In addition, both ethnographic accounts (Molefe, 2015;Rifkin, 2015b) and recent experimental works (Hodgskiss, 2010;Rifkin, 2012;Wadley, 2005) have resulted in more functional interpretations of ochre uses, for example, as an insect repellent (Rifkin, 2015a), sunscreen (Rifkin et al., 2015), as a hide-tanning agent (Rifkin, 2011) or as an ingredient in hafting mastics (Lombard, 2006(Lombard, , 2007Wadley et al., 2004;Zipkin et al., 2014).

| Blombos Cave and its Middle Stone Age ochre assemblage
Blombos Cave is located ca. 300 km east of Cape Town on the southern coast of South Africa. It contains more than 3 m of laminated deposits dated to ca. 101-70 ka (Henshilwood et al., 2011;Jacobs et al., 2020;Tribolo et al., 2006). The MSA sequence has been divided into four occupation phases, of which the upper two-M1 and M2 upper-are associated with the Still Bay technocomplex and dated to ca. 76-71 ka (Jacobs et al., 2020).
Around 8000 pieces of primarily red ochre are reported from the MSA levels, of which 1,500 are >10 mm in length (Henshilwood, Sealy, et al., 2001;Watts, 2009). The majority of the anthropogenically modified ochres (n = 1534) are deep, saturated shades of red (Watts, 2009). Although most of the ochres recovered from BBC are associated with the upper M3 occupation levels dated to ca. 84 ka, a considerable amount of modified ochre (n = 254) is also found in the M1 and M2 upper phases, that is, the Still Bay phase. Two modified ochre pieces from the M1 phase display cross-hatched scorings in combination with parallel incised lines (Henshilwood et al., 2002). Henshilwood et al. (2009) argue that the geometric arrangement of the score marks on these pieces represents engravings that may constitute a premeditated pattern or design, as both pieces appear to have been subjected to highly consistent scoring pressure, indicating that the lines were produced in a precise and controlled manner.
A range of other artefacts associated with ochre is also reported from BBC. Henshilwood et al. (2018) report a silcrete flake displaying a cross-hatched pattern drawn on with an ochre crayon from the Still Bay levels. Several ochre-stained and perforated marine shell beads are documented from the same levels (d 'Errico et al., 2005'Errico et al., , 2015Henshilwood et al., 2004;Vanhaeren et al., 2013). Additionally, excavators have also recovered an in situ ochre toolkit consisting of ground ochre powder, seal bone and charcoal mixed into a liquid compound and kept inside two abalone shell containers from the lowermost level of the cave, dated to ca. 101 ka (Henshilwood et al., 2011).

| ANALYTICAL SCOPE
In this study, we characterise the trapped ochre fragment in the following contexts: (1) Archaeological field context-by visually and spatially contextualising the ochre within its original location in the cave and its associated occupational phase; (2) Archaeological microcontext-by analytically and spatially situating the ochre within the microlaminated occupation deposits that physically surround it; (3) Physical properties-by reconstructing its size, shape, texture, internal structure as well as its elemental and mineralogical composition; (4) Anthropogenic surface modifications-by documenting and analysing the occurrence, location and physical properties of scoring marks and striations.
For the interpretation of the trapped ochre fragment, two conventionally recovered and previously described ochre pieces (SAM 8937 and SAM 8938) were selected for morphometric comparison (Henshilwood et al., 2002). These two reference ochre pieces are from the same occupation phase as the trapped ochre (Still Bay) and were recovered only 1.3 m away from the ochre in the block sample ( Figure S2). Both reference ochre pieces were ground and facetted before being scored (Henshilwood et al., 2002) and Henshilwood et al. (2009) also report-besides the occurrence of engravings-the presence of microstriations and striations on their surfaces.

| Archaeological field context (macroscopic)
The MSA occupation deposit (M2 upper) associated with the block sample containing the trapped ochre (BBC-13-16) was first excavated in 2011 and then again in 2019. We used field notes and photographs taken during these field seasons, in combination with image-based 3D modelling techniques (Unhammer, 2016), to reconstruct and characterise the sedimentary matrix and depositional setting in which the ochre piece was recovered.

| Micromorphology and archaeological microcontext
The block sample (BBC-13-16) was collected from the exposed southern section wall of BBC ( Figure 2). It was stabilised by plaster bandage, before being transported to the University of Tübingen, Germany, where it was impregnated with resin under vacuum, in a 7:3 mixture ratio of unpromoted polyester resin (Viscovoss N 55S) and styrene, in addition to 5-ml/L hardener (methyl ethyl ketone peroxide). The hardened block sample was sliced into four slabs ( Figure 3a), one of which was selected for thin sectioning (60 × 90 mm, 30-μm thickness). It should be noted that during the slicing of the impregnated block sample, some of its original thickness, including the trapped ochre piece, was removed by the saw blade (>10 mm). By applying thin section-wide documentation procedures described by Haaland et al. (2019), and by following general (Courty et al., 1989;Stoops, 2003) and site-specific descriptive protocols (Haaland, 2018), we characterise the sedimentary composition and microstratigraphic configurations.

| Micro-CT scan and 3D surface reconstruction
Each of the block sample slabs (four in total) was individually scanned with micro-CT at the General Electric office in Stuttgart with a Phoenix v|tome|x m CT scanner (General Electric) at 250 kV, 220 mA and an isotropic resolution of 0.0397 mm. Segmentation of each micromorphological slab (n = 3) was performed automatically using thresholds optimised for best segmentation results in Avizo 8.1 (Thermo Fisher Scientific-FEI). The removal of the pervasive nonochre material, such as attached quartz grains, finer calcareous sediments and resin, was done manually. A tri-dimensional surface (i.e., a cloud of 3D points connected by triangles) was obtained from each segmented volume by allowing for maximal and unconstrained smoothing (level 9 in the function "Surface Gen"). Finally, the ochre pieces that were inside the micromorphology block were combined and manually aligned to restore the overall morphology of the original piece before it was cut.
We also conducted high-resolution micro-CT scanning on the two reference ochres from BBC (SAM 8937 and SAM 8938) (Henshilwood et al., 2002). These scans took place at the Microfocus CT Laboratory (Evolutionary Studies Institute), the University of the Witwatersrand, South Africa. Scanning was performed with a Nikon XTH 225/320 LC with a Perkin Elmer flat panel detector at 90 kV, 220 mA and isotropic resolutions of 0.0346 and 0.0498 mm. From these scans, we created two digital 3D models that were generated and rendered in the same way as the trapped ochre piece, thus providing us with a comparative data set.

| Microanalytical techniques
Micro-XRF analyses were conducted on the thin section and corresponding slab, enabling the creation of elemental distribution maps using a Bruker M4 Tornado benchtop instrument.
Analyses occurred under full vacuum (20 mbar), with no filters, an Rh tube voltage of 50 kV and 600 mA current, and a detection range of 0−40 keV. The average spot size is 20 µm; pixel spacing and dwell time varied from 8 to 60 µm and 10 to 300 ms, respectively, depending on the mapping area. We used one of the two silicon drift detectors in the system that is optimised for quantitative analyses, calibrated the instrument using a zirconium standard and checked the instrument function using measurements on internal steel, glass and copper standards. The reported elements (see Table 1) were selected by manually observing the presence or absence of K-and L-lines in the average spectrum generated over the entire mapping area and on the maximum pixel spectrum generated on maps of the ochre only.
The elements were subject to a modified version of the Bruker "oxides"-a quantification method that presents the results as wt% oxides. The formulae listed in parentheses in Table 1 after the elements indicate the oxides that were used in the wt% and norm. wt % calculations. Rh, which is present in the instrument itself, is used only in the deconvolution. In this study, we quantified the elemental composition of seven ochre fragments (originally belonging to a single ochre artefact) by conducting deconvolution on average spectra generated from an area corresponding to each of their surfaces (see outlines of measured fragments 1−7 in Figure S5). Quantitative results were generated from areas that met or exceeded 300 s of analytical time. The elemental mapping allowed a visual observation of the distribution of elements within the fragments, and additionally, transects were generated from the maps to evaluate minor fluctuations in element abundance along different planes.
For the mineralogical characterisation, we used FTIR on a few micrograms of ochre powder, obtained by scraping the internal edge of the ochre within the already cut chip. The instrument used was a Cary 660 bench FTIR spectrometer (Agilent Technologies). The measurements were conducted in both transmission mode and in attenuated total reflectance (ATR) mode (64 scans at 4 cm −1 resolution), allowing us to define absorption peaks in the frequency area of 4400−400 cm −1 . For transmission mode measurements, the ochre sample was ground into a powder using an agate mortar and pestle. Approximately 0.1 mg of sample was mixed with 0.5 mg of KBr powder and then pressed into 7-mm pellets using a hand press. The FTIR spectra were processed and analysed using the Agilent Resolutions Pro software and compared with in-house references and other available reference libraries (Lafuente et al., 2016;Weiner, 2010). Spectra were also collected directly off the chip and petrographic thin section using the FTIR microscope (Cary 610) attached to the same instrument in reflectance mode and using a diamond crystal ATR accessory (Agilent Pike Gladi-ATR Vision). Raman spectra were obtained using a confocal Raman spectrometer (Horiba Jobin-Yvon Labram-HR) attached to a petrographic microscope and collected in the spectral range between 150 and 1500 cm −1 using 514-nm laser wavelength (25 s exposure time and 25 acquisitions).
The measurements were collected through a 50× objective on a confined area of the thin section containing the ochre (c. 5 µm 2 ).

| Digital reconstruction and quantitative characterisation of ochre surfaces
Although there are few high-resolution investigations of surface modifications on ochre (but see Rosso et al., 2017), a quantitative cut mark analysis has a long history within faunal taphonomy (Binford, 1981;Lartet & Christy, 1875;Shipman, 1981). Particularly relevant for this study is the development of morphometrical analysis of cutmarks using high-resolution 3D digital recording methods, HAALAND ET AL.    Note: Seven individual fragments were measured, with outlines given in Figure 4b. The formulae listed in parentheses after the elements indicate the oxides that were used in the wt% and norm. Wt% calculations (see Tables S2−S8 for more detailed results). Abbreviations: m-XRF, micro-X-ray fluorescence.
Here, we adopt a similar approach that allows us to physically measure and compare the modifications on the surfaces of the trapped ochre fragment with those found on the two reference ochre pieces. Specifically, we converted the 3D ochre models made from the micro-CT scans to digital elevation models (DEMs). We then imported these DEMs into mapping software (Surfer 16) and made them into digital microtopographic surfaces, allowing us to conduct metric measurements and to create profile crosscuts (i.e., red lines in

| Archaeological microcontext
The upper M2 microstratigraphy is characterised by a complex sequence of unconsolidated sandy deposits dominated by silt to sand- between their surfaces. We estimate the minimum original size of the ochre piece to be ca. 5.9 × 3.2 × 3.7 cm (see Table 2 and Figure S1). The elongated morphology of the trapped ochre is broadly rectangular, having three flat sides (Sides A and C and the bottom part), whereas the top part displays a convex profile shape.
This morphology appears to be uniform throughout the different sections of the piece ( Figures S3 and S4).

| Elemental and mineralogical characterisation
Micro-XRF elemental distribution maps were produced from the petrographic thin section (Figure 13a). These maps showed the relative abundance of elements in the ochre piece, relative to the surrounding sedimentary matrix, and within the piece itself. Compared with the sedimentary matrix which is rich in Ca, Si and P, the  (Table 1).
Representative profile shapes of each modification type are provided in Figure 16a-g. A boxplot showing the numerical spread and variation within and between each measurement category (angle, height and width) by the modification type is shown in Figure 16h (see Table 3 for summary statistics and Table S1 for raw data). It should be noted that the y-axis (i.e., the height of the profile) in Figure 16a-g has been amplified by an elevation factor of three (z = 3×) to make the profile shapes more distinguishable. Although this uniform data set adjustment ensures that our comparative analyses are internally consistent, the values presented in Table 3 do not represent absolute, real-world values.
The profile shapes and boxplots in Figure 16 allow for visual and quantitative characterisation of the relative differences between each modification category. Microstriations represent small and shallow surface modifications that are <0.1 mm in width (Table 3). To further compare the main modification types and the two engravings, we used multivariate statistics. Figure 17a shows a principal component analysis (PCA) of profile shape measurements (n = 64) by the type of modification. The first two principal components confirm that the three main modification types and the two sub-types generally form consistent and distinct shape-based groups.
We also compared the profile shape measurements of the two crosshatched engravings with the multivariate mean of the predefined modification types using a canonical discriminant function analysis (CDA) (Figure 17b). Qualitatively, the engravings of SAM8937 were defined as shallow , and indeed, their profile shape measurements plot partially between the scoring and the striation categories and partially outside them (blue crosses in Figure 17b). Some of the shallow engraving measurements fall outside any of the categories, having slightly larger W1 and W2 values than the striations, and slightly wider angle than scoring. The engravings of SAM8938 were reported to be deeper and most likely made from repetitive and controlled hand motions. Again, the multivariate plots of their profile shape measurements support this qualitative assessment, as they regularly plot between the scoring and the deep scoring category (red circles in Figure 17b).
From these results, we conclude that our reference data set enables us to characterise and classify surface modifications as accurately and consistently as conventional qualitative assessments. Ultimately, this result enables us to apply our quantitative classification scheme on surface modifications we cannot physically observe, such as those present on the digitally rendered surface of the trapped ochre piece.  Finally, we evaluated the spatial accuracy of 3D reconstructed ochre surface modifications by comparing their location and orientation with high-resolution edge profiles inferred directly from F I G U R E 11 A combined three-dimensional (3D) reconstruction of the trapped ochre fragment inside block sample BBC-13-16. Each fragment (I, II and III) and each side (A, B, C, D, top and bottom) have been marked for later reference (see Figure S3 for 3D renderings of each individual fragment) [Color figure can be viewed at wileyonlinelibrary.com] T A B L E 2 Dimensions (mm) and provenance of the trapped ochre fragment (BBC-13-16-O1) compared with two previously reported modified ochre pieces (Henshilwood et al., 2002)  the physically trapped ochre (Figure 8a−c). Three parallel score marks were specifically selected for this procedure (lines 4-6 in Figure 19c). The results show that the score marks identified on the digitally rendered ochre piece align well with microtopographic profiles made from the physical ochre profiles (Figure 8d).

| Archaeological and depositional setting
The BBC-13-16 block sample was collected ca. 1 m behind the cave drip line, in a part of the cave that functioned as a central activity area throughout all MSA occupation phases (Haaland, 2018). This was also the case during the phase in which the trapped ochre was deposited, that is, the Still Bay phase. In general, the Still Bay is archaeologically associated with the earliest, but infrequent, occurrences of bifacially worked stone points (Villa et al., 2009), polished bone points d'Errico & Henshilwood, 2007), perforated marine shell beads (d 'Errico et al., 2005;Henshilwood et al., 2004;Vanhaeren et al., 2013) and at least two engraved ochre pieces ; however, numerous other modified and nonmodified ochre artefacts are also reported (Henshilwood, Sealy, et al., 2001;Henshilwood et al., 2002;Watts, 2009 (Haaland, 2018). This interpretation is reinforced by 3D reconstruction of the occupation surfaces surrounding the trapped ochre (Figure 8), which show the presences of large amounts of lithic debris, faunal remains and shellfish fragments.
In the thin section (Figures 9 and 10 Table 1 for XRF overview data and Tables S2−S8 for detailed  suggests that these sediments may still have been moderately reworked and disturbed through dumping or sweeping activities (Goldberg et al., 2009;Miller et al., 2013).
Finally, the trapped ochre piece was found in close proximity to another large ochre piece, recovered within the same layer, some 20 cm apart (Figure 8c). It is, thus, possible that both ochre pieces were in use during the same occupation event, and that both were intentionally left behind or discarded.
Unlike the dense and trampled occupation deposits that the ochre is resting on, the sedimentary matrix that surrounds and covers it contains larger and less fragmented pieces of shellfish and bone that show less sign of compaction (Figure 10b). This may indicate that the ochre piece was left at the site during an

| Evaluating the mineralogical and elemental analyses
The geochemical methods employed in this paper were selected for several reasons. First, we chose methods that would yield both elemental and molecular information, as similar approaches are typically employed in conventional ochre studies (e.g., Bouillot et al., 2017;Román et al., 2019;Wojcieszak & Wadley, 2019;Zipkin et al., 2017).
Second, we aimed to achieve nondestructive or minimally destructive sampling. Although this may seem counterintuitive-as one could argue that the ochre piece has already been destroyed-there is an increasing Elemental analyses were conducted here using micro-XRF. The choice of instrument reflected its availability and spacious sample chamber that can accommodate large blocks. A scanning electron microscope equipped with an energy-dispersive X-ray detector or an electron microprobe with wavelength-dispersive X-ray detectors is also an appropriate technique for these types of samples and will produce similar information. Likewise, we selected micro-FTIR and micro-Raman for mineralogical analyses due to instrument access; micro-X-ray diffraction would also be a comparable choice.
Analyses using micro-FTIR and micro-Raman are nondestructive to the petrographic thin section and slab, and both techniques yielded suitable data on the trapped ochre piece. The lack of resin peaks in the FTIR spectra suggests that there was little penetration of the embedding medium into the ochre piece, which may help to explain the success of the Raman analyses. Micro-Raman analyses have been previously attempted on micromorphological samples (Goldberg, personal communication;Mentzer, 2011), but strong fluorescence of the resin likely explains the lack of published applications (Beyssac et al., 2002;Dupin et al., 2019). Lastly, both FTIR and Raman have been successfully employed in southern African MSA ochre studies (Dayet et al., 2013;Hodgskiss, 2012;Moyo et al., 2016), and our results are directly comparable to these studies.

F I G U R E 15
The micro-Raman spectra of the trapped ochre piece, as measured directly on the uncovered thin section, compared with the reference Raman spectra of haematite from the RRUFF project (Lafuente et al., 2016) and the RDRS project (Buzgar, 2009) There has been one previous study on Blombos ochre using analytical techniques by Moyo et al. (2016). In this study, Moyo and authors reported the most common elements found in some 68 ochre pieces detected by ED-XRF, which were Al, Si, K, Ca, Ti and Fe. This corresponds well with the elemental composition of the trapped ochre piece, which we measured through m-XRF. The lack of Mn in the trapped ochre sample is not unusual for Blombos. Whereas Mn-rich ochre artefacts have been identified, they only occur in small numbers and they do not occur consistently throughout the MSA sequence (Moyo et al., 2016).
The nearby Bokkeveld shale formations (Danchin, 1970), which Moyo et al. (2016) reports may have offered sources of ochre during the MSA inhabitants of BBC, also contains low traces of Mn.

| Size, shape and physical properties
The trapped ochre (BBC-13-16-01) is one of the largest modified ochre pieces found in Blombos (Henshilwood et al., 2002;Henshilwood et al., 2009). It displays multiple flattened and faceted surfaces that are typical of intensive grinding, and the rounding of the top portion may also be the result of grinding, followed by rubbing or smoothing (Hodgskiss, 2010). The presence of these mor- It should be noted that the y-axis (i.e., the height of the profiles in (a-g) has been amplified by an elevation factor of three (z = 3×) to make the profile shapes more distinguishable F I G U R E 17 (a) Principle component analysis of profile shape measurements (angle, height and width; n = 64) of surface modifications on ochre by the type of modification (see Figure 15 and Table S1). (b) A canonical plot for a linear discriminant analysis of profile shape measurement (using quadratic, different covariances) showing the multivariate mean of each modification type (inner circle = 95% confidence level ellipse, outer circle = 50% class contour). Measurements of the two cross-hatched engravings are shown by blue crosses (thin engravings on SAM 8937, n = 21) and red circles (thicker engravings on SAM 8938, n = 28) [Color figure can be viewed at wileyonlinelibrary.com] 2010). Ochres with a hardness ranging from 2 to 5 can produce colour streaks. As such, we may envision that the trapped ochre would have been an ideal candidate for creating pigment powder that exhibited a deep red colour (Watts, 2002), similar to numerous ochre pieces previously documented from BBC . Combined with its homogenous fine-grained texture ( Figure 12), its elemental and mineralogical composition of haematite, kaolinite and quartz (Figures 13, 14, and 15) and its fine, internal laminations ( Figure S7), the trapped ochre piece seems to represent a Fe-rich, fine-grained material, such as shale or mudstone. Both are common in the Bokkeveld Group outcrops located some 20 km from the site, and this has previously been suggested as the most likely source for many of the BBC ochres Watts, 2009).
The micro-CT images revealed several internal weakness planes that run internally across the entire ochre piece ( Figure S4).
F I G U R E 18 (a) Side C of the trapped ochre piece   iron-rich ochre fragments from the surrounding quartz-rich sandy matrix within the digitally reconstructed matrix. Second, the ochre piece was accidentally cut into three main slabs by the rock-saw operator during the initial processing stage, which meant that each slab had to be individually scanned and digitally processed before being aligned and made into a single model. Third, as we wanted to document even the smallest of surface marks (e.g., microstriations), the micro-CT scans were conducted at a very high resolution (the raw scanning volumes amounted to very large files of ca. 22 GB in total). To balance quality and efficiency, we used a combination of low-resolution full-scale models and high-resolution close-up views to document the occurrence, location and types of surface modifications present on the trapped ochre piece.
6.5 | Evaluating the microtopographic classification of ochre surface modifications The reference data set we used in this paper relied on a limited number of already known modified ochre pieces (n = 2). As such, these data cannot fully account for all types of anthropogenic ochre modifications. However, considering the consistent distribution patterns of morphometric measurements within the PCA and CDA plots ( Figure 17), we believe our reference data set to be a reasonable representative for at least three types of generic surface mod- 6.6 | Anthropogenic surface modifications present on the trapped ochre and behavioural implications The artificially flattened and faceted topography suggests that all sides were once intentionally ground, and as such the entire morphology of the trapped ochre is a product of human modification.
Because of the fragmentation and the crosscuts caused by the rock saw blade, it is challenging to assess the precise location and distribution of all types of surface modifications located on the trapped ochre. Nonetheless, two of the flattened sides (A and C on Figure 11) have distinct striation and score marks that run parallel with the elongated axis of the piece. Our 3D reconstruction and profile shape analysis show that several of these surface modifications can be classified as different types of score marks. Side A, which exhibits between five and eight parallel lines ( Figure 20) can all be classified as either scorings or deep scorings, making them comparable in size and shape to cross-hatched pattern incisions on the SAM 8839 ( Figure 6). Side C contains at least five scorings (- Figure 18). The score marks on both side A and C are all aligned with the general, elongated axis of the ochre piece itself, and their paths seem to cross each other randomly. On the basis of these observations, it is likely that the score marks on sides A and C of the trapped ochre were intentionally made and represent incisions produced by a sharp object.
Do the lines on Side A or Side C of the trapped ochre piece represent instances of intentional engraving, that is, some type of overarching design? According to Henshilwood et al. (2009), examples of engravings include, but are not limited to, multiple score marks that constitute a deliberately arranged pattern which would have required a specific and controlled hand motion and would, thus, have been incompatible with effective powder production. They furthermore argue that engravings typically comprise similar-sized, juxtaposed score marks that exhibit uniform cross-sections, indicating that a constant pressure was applied during the incision process of all the lines and that the lines were produced by the same tool during a single session. Following the criteria of Henshilwood et al. (2009), the variety of score marks on Side C do not qualify as engravings. The parallel lines on Side A, on the contrary, appear to be made with the same tool and pressure, they are uniformly distributed, and they appear to be deliberately aligned with each other and with the edge of the piece. Consequently, they bear a clear resemblance to previously documented engravings from BBC , and would most likely have been defined as such if a conventional, qualitative investigation had been possible.

| A brief discussion on ochre modification terminology
Although numerous studies of African ochre assemblages focus on identifying anthropogenic use traces on ochre pieces (Bernatchez, 2012;Bouillot et al., 2017;Dayet et al., 2016Dayet et al., , 2017de la Peña et al., 2019;Henshilwood et al., 2002Henshilwood et al., , 2009Hodgskiss & Wadley, 2017;Hodgskiss, 2012Hodgskiss, , 2013Rosso et al., 2017;Watts et al., 2016;Watts, 1999Watts, , 2002Watts, , 2009Watts, , 2010, there is currently no consensus on terminology for specific types of ochre modifications. Though certain terms are, in general, agreed upon (i.e., striation, crayon and scoring), in many cases, the researchers often use terms adapted to the local context and site-specific ochre assemblage. For example, both Watts A study by Hodgskiss (2010) sought to rectify some of these discrepancies by identifying the specific behaviours or actions responsible for certain types of modifications seen on experimentally modified ochre pieces. Following a similar rationale, Rosso et al.
(2017) explored the effect that different grindstone textures had on the coarseness of the ochre powder being produced. However, both studies focused more on linking ochre processing actions to specific types of surface modifications. Less emphasis was placed on quantitative documentation and defining the observed modifications, in terms of their 3D morphology and metric size.
Within this context, our paper may serve as a case study that demonstrates both the need and the benefits of quantitatively characterising and classifying ochre surface modifications in a standardised and consistent way. Profile shape measurements on ochre can easily be conducted on conventionally recovered pieces. Microtopographic data sets can be acquired not only through micro-CT scanning, but also through alternative methods such as confocal microscopy or microphotogrammetry. Consequently, we regard our profile shape classification scheme to function as a proof of concept that can encourage others to adopt similar documentation strategies.
This will improve intersite comparisons of ochre modifications and may thus reduce confusion in terms of ambiguous terminology.
Although we acknowledge that our ochre reference samples are locally constrained within the scope of this paper, we also believe that the observations and patterns presented here should be further tested and evaluated by controlled experiments aimed towards creating a classification scheme that is both quantitatively defined and behaviourally informed. Indeed, several of the authors of this manuscript are currently involved in a more comprehensive experimental study that focuses exclusively on linking physical modification variables, for example, ochre hardness, scoring pressure and type of implement used, to their microtopographic signatures.

| CONCLUSION
Our main objective in this study was to reconstruct the archaeological context, texture, composition and morphology of the trapped ochre piece and to document and characterise any possible surface modifications. For these purposes, we combined 3D field documentation, micromorphology and high-resolution micro-CT scanning with microanalytical techniques applied directly on the micromorphological slab and thin sections. The results of our investigation are as follows: • The trapped ochre piece was deposited within a moderately reworked occupation deposit of the upper M2 phase (dated to c. 74 ka), alongside other artefacts such as lithics, bone fragments and shellfish • It was a minimum of 5.9 × 3.2 × 3.7 cm in size, making it one of the largest modified MSA ochre pieces recovered from BBC • It had a fine-grained texture and a clay and iron-rich mineralogical composition • It was intentionally facetted on all sides, most likely due to intensive grinding • It had multiple scoring marks, some of which qualify for the definition of engravings Our results demonstrate that a block sample-based study of a trapped archaeological artefact allows for an unusually in-depth, multiscalar and multiproxy study, enabling us to assess its depositional and microcontextual archaeological setting, its morphology and microtopography, its texture and its geochemical properties at the same time. In our specific case, the micro-CT scanning also facilitated a quantitative, morphometric profile analysis that permitted us to conduct quantitative comparisons of the surface modifications on the trapped ochre with those present on reference ochres. Even though the scope of our case study is limited to a single ochre piece, this development of a quantitative classification scheme for ochre surface modifications represents a novel attempt to objectively describe, compare and classify ochre typologies that traditionally have been only qualitatively characterised.
Although ochre materials do not occur in every archaeological context, we believe that the analytical approach presented here can be adapted and applied to any number of other materials, which may occur within a micromorphological block sample, such as bones, figurines and metallic objects, to name a few. Still, we must emphasise that micromorphological block sampling should not be avoided out of the concern of accidentally collecting significant archaeological artefacts. Through careful sampling and systematic assessment of the local field context, the risk of collecting such artefacts can be significantly reduced. In the unlikely event that complete archaeological artefacts do end up in a block sample, however, our study shows that information is neither lost nor destroyed. Rather these trapped artefacts can be robustly reconstructed and innovatively documented at a high resolution and through a wide range of spatially, physically and contextually sensitive analytical methods. Ultimately, we envision that this type of multiscalar and multidisciplinary framework may prove beneficial in the study of conventional archaeological artefacts-and not just those that are trapped.

ACKNOWLEDGEMENTS
Financial support for the analysis of the block samples was provided