Animal Economies and Islamic Conversion in Eastern Ethiopia: Zooarchaeological Analyses from Harlaa, Harar and Ganda Harla

Excavations at three urban sites, Harlaa, Harar, and Ganda Harla, in eastern Ethiopia have recovered substantial assemblages of faunal remains. These, the first to be analysed from Islamic contexts in the country, were studied to reconstruct animal economies, and to assess if it was possible to identify Islamic conversion or the presence of Muslims in archaeological contexts through examining butchery practices and diet via the species present. Differences in animal economies between the sites in, for example, management strategies, use of animals for traction, and presence of imported marine fish, infers the development of different traditions. However, conversion to Islam was evident, and although issues of non-observance, mixed communities, and dietary eclecticism have to be acknowledged, the appearance of a similar range of butchery techniques suggests these were linked with the appearance of Muslim traders, and subsequent spread of Islam.


Introduction
Excavations completed as part of the Becoming Muslim project (694254 ERC-2015-Adg) have recovered substantial assemblages of faunal remains from sites in Harlaa (2017)(2018)(2019), Harar (2014Harar ( , 2018, and Ganda Harla (2014). These are the first to be published from Islamic sites in Ethiopia, and one of only a few assemblages to have been analysed from sites that are contemporaneous with the better-studied Aksumite (c. 1st-7th centuries AD) and post-Aksumite contexts in the north of Ethiopia and neighboring Eritrea (e.g. Cain 2000;Lesur et al. 2007;Chaix 2013;González-Ruibal et al. 2014;Woldekiros & D' Andrea 2017). The zooarchaeological analyses were completed with the aim of reconstructing animal economies (herd management strategies, sex proportions of domesticates, body portion distributions, and pathological bone modifications), and assessing if it was possible to identify Islamic conversion or the presence of Muslims in archaeological contexts through examining butchery practices and diet via the species present. Both indicators were useful in recognizing Islam, but non-observance is a factor which has to be acknowledged. Moreover, significant differences were evident between the three sites illustrating variability in animal economies and the development of different traditions over time within eastern Ethiopia.
Answering the questions posed was facilitated by the fact that the faunal material from all three sites demonstrates excellent preservation of both dense and less dense elements. Costal cartilage was recorded in a majority of excavated contexts and foetal, neonatal and infant remains were also recovered from across the sites (Supplementary Tables S1-S28). A reasonably low degree of pre-and postdepositional fragmentation of material was also evident. The systematic dry-sieving of all contexts that was employed allowed the recovery of both small elements and remains of small animals, including birds, fish, and rodents.

Harlaa (9°29′10.22″N, 41°54′36.96″E)
The archaeological site of Harlaa is located at 1700 m ASL on the edge of the main fault escarpment of the southern Afar margin underneath the modern Oromo village of Ganda Biyo on the Dire Dawa to Dengego road. The accepted name for the archaeological site is Harlaa which is related to the common appellation 'Harla' given to ruined stone-built towns and funerary monuments in the region, whose origins are ascribed by the Oromo to a legendary ancient people of giant status (Chekroun et al. 2011: 79), and who occupied the region before the Oromo arrived (Joussaume & Joussaume 1972: 22), beginning in the mid-16th century (all dates are AD unless otherwise specified). Harlaa is situated approximately 40 km north-west of Harar and 15 km southeast of Dire Dawa (Fig. 1). Prior to the start of the current investigation of the site, previous research had consisted of limited survey and surface collections (e.g. Patassini 2006). Harlaa is a large urban centre (Fig. 2) covering a maximum area of approximately 500 m east to west by 900 m north to south, excluding outlying cemeteries (Khalaf & Insoll 2019). It is composed of several elements including a central settlement area, workshops, at least three early mosques, wells, lengths of fortification wall, and cemeteries to the north, east, and west.
Since 2015, excavations have been completed in a mosque (Area A), workshop complex (Area B, except labelled A in 2016), cemeteries (Areas C and D), a house with associated industrial/kitchen facility (Area E), and part of an extensive building complex, which included a defensive wall, and what may have been a reception hall (Area F). Faunal assemblages were recovered from areas B, E, and F. The chronology at Harlaa predates both Harar and Ganda Harlaa. Area B has provided the longest  11th to mid-13th centuries AD 3 Late 12th to late 13th centuries AD 4 Mid/late 13th to early 14th centuries AD 5 Late 13th to 14th centuries AD The 2017 to 2019 excavations in Area B recovered 33,184 faunal remains totaling 90,422.5 g, of which 8063 (24.3%, by weight 63323.8g or 70.0%) were identifiable. Of these, 2655 (32.9%, by weight 38599.1 g or 61.0%) were identifiable to genus level. A further 3897 (48.3%, by weight 20070.6 g or 31.7%) fragments were identifiable to family level and an additional 1450 (17.9%, by weight 4625.0 g or 7.3%) identifiable only to element. The 2018 and 2019 excavations of Area E recovered 13,155 faunal remains totaling 23060.6 g. Of these, 612 were identifiable to genus level (4.6%, by weight 7369.4 g or 31.9%), 1079 identifiable to family level (8.2%, by weight 4956.6 g or 21.5%) and 399 (3.0%, by weight 895.5 g or 3.9%) to element and animal size class. The 2019 excavations in Area F recovered 1847 fragments totaling 2533.0 g. From this sample 93 fragments were identifiable to genus level (5.0%, by weight 952.7 g or 37.6%), 125 (6.8%, by weight 425.4 g or 16.8%) were identifiable to family level and 49 (2.6%, by weight 74.8 g or 3.0%) were identifiable only to element and animal size class. Unidentifiable material from these excavations included small fragments of cranium or long bone shaft which could not be further identified to element. In comparison to the identifiable elements, the 25,121 unidentifiable fragments from Area B averaged 1.08 g (compared with 7.85 g for identifiable material), the 11,037 unidentifiable fragments from Area E averaged 0.89 g (compared with 6.34 g for identifiable fragments) and the 1580 unidentifiable fragments from Area F averaged 0.68g (compared with 5.44g for identifiable remains). In the majority, these comprised fragments of appendicular cortical bone (shaft fragments), although the remainder of the skeleton (fragments from the cranium and axial portions) were also well represented. Discussions of fragments identifiable to the level of element, family and genus are hereafter discussed in NISP (Number of Identified Specimens) unless otherwise stated. Further detail as to the distribution of unidentifiable fragments by fragment type and animal size class is available in the Supplementary Table S1.

Harar (9°18′33.21″N, 42°8′15.84″E)
Harar is also situated in the Somali Plateau, at 1900 m ASL, but in a more heavily vegetated landscape (Khalaf & Insoll 2019). At the core of Harar is a historic city, surrounded by a wall, the djugel. This is built of locally quarried calcareous tuff (Hashi stone) mortared with mud and wooden reinforcements (Ahmed 1990: 321). It encompasses an area of c. 1000 × 800 m and is accessed by five gates, with corresponding quarters (Horton 1994: 195). The walled city contains within it approximately 2000 houses, 82 mosques, and numerous saints' tombs and shrines (CIRPS & Harari People National Regional State 2003), the result of important urban development over many centuries (Insoll 2017).
Excavations at nine locations in and immediately adjacent to the walled city (five mosques, one shrine, two settlement areas, one blacksmithing location) (Fig. 3), have provided eight AMS radiocarbon dates of between the 15th and 19th/early 20th centuries (cf. Insoll 2017; Insoll and Zekaria 2019). The excavations indicate that occupation in Harar post-dates the late 15th century, and until evidence to the contrary is found, it is suggested that the city and its mosques date from this era and were linked with the establishment of Harar as the capital of the Sultanate of Adal (c. 1415 to 1577) (Insoll 2017;Insoll and Zekaria 2019). Prior to this the Harari, likely in the form of the 'Harla' , were elsewhere, possibly at Harlaa or one or more of the largely uninvestigated abandoned stone town sites, such as Ganda Harla, that are found across the eastern Harar Plateau and the Chercher Mountains (cf. Insoll 2017: 209-210). Although Harar functioned as a trade centre that connected the eastern Ethiopian highlands, arid lowlands, and the Gulf of Aden, the situation was very different to Harlaa as few luxury materials have been recovered. These comprise four glazed ceramic sherds of probable Middle Eastern Origin, one sherd of Chinese blue and white ware of 18th or 19th century date and a single fragment from a green glass bottle or flask neck (Insoll 2017: 207;Insoll and Zekaria 2019).
For the purposes of this discussion, the faunal remains from all the units in Harar are combined rather than presented as separate assemblages. This does not bias the analysis as there were no meaningful disparities in the quantities found or species present. The excavations at Harar recovered 887 bone fragments totaling 4974.8g. Of these, 173 (19.5%, by weight 1859.3 g or 37.4%) were identifiable to genus level, 142 (16.0%, by weight 1326.4 g or FIGURE 3 The location of the excavations in Harar (prepared by N. Khalaf).

Taxonomic Distributions
Identified faunal remains were categorised according to genus and species, where possible. Where specific taxonomic identification was not possible, elements were classified according to the most precise possible broader taxonomic category, such as small bovid, bird (aves), small carnivore, or medium mammal. Taxonomic identifications were made using Boessneck et al. (1963), Olsen (1960, Walker (1986), Zeder & Lapham (2010), and Zeder & Pilaar (2010) as well as a digital reference collection. Each element was identified either to taxon or taxonomic category with further information recorded as to element side, percentage of completeness, age or sex information where present, breakage patterns, burning, weathering, gnawing, butchery evidence (such as cutmarks), and other taphonomic information. Identified elements are given here as NISP (Number of Identified Specimens). Comparisons of the representation of elements for the most common mammals in each faunal sample indicates no significant differences in the representation of bodily portions either through time or between the sites (see Supplementary Table S1). In all cases, the distribution of elements suggests the presence of animals as complete carcasses.
The majority of fauna from all three sites come from domestic taxa (Table 4). For those remains which could be identified to genus level, domesticates represent between 59-100% of NISP with the majority of samples containing c. 60-77% domestic fauna (Table 5). At Harlaa, the most common domesticate is the goat (Capra hircus) at c. 35-40% whereas at Harar and Ganda Harla it is cattle (Bos taurus/indicus) at 45-64% (Tables 4 and 5). The sole exception is the very small sample from Phase 1 of Area B at Harlaa, where the camel is the most common domesticate (26.3%). The second most common domesticate is either cattle (Harlaa) or goat (Harar and Ganda Harla). Sheep at Harlaa (excepting the small sample from Phase 1) are the third most common domesticate followed by transport animals. This situation is reversed at Harar and Ganda Harlaa where transport livestock (camel, donkey and horse) combined are more common than either sheep or goats (Tables 5 and 6). The representation of transport animals in different areas of Harlaa is uneven, with an overall decline in Area B through time and a greatly reduced representation in Areas E and F (6.3% and 7.5%) compared with the contemporaneous Phase 2 sample from Area B (14.9%). Transport animals across all contexts are dominated by equids (donkey and horse), with the decrease in transport animals in Area B largely due to the decreasing presence of camels in later phases ( Table 6).
Remains of domestic chicken (Gallus gallus domesticus) were recovered from Harlaa and Harar but are absent from Ganda Harla (as are all bird bones). These are systematically present across all contexts at Harlaa except the small sample from phase 1, Area B, and their representation is fairly consistent at c. 3-7% at Harlaa and 2.3% at Harar. Males appear to be infrequent as adults but are represented as subadult individuals, suggesting that extraneous roosters were culled prior to adulthood. Domestic guinea fowl (Numida meleagris) is present only at Harlaa, and is seen only in Areas E and F and phases 2 and 3 of Area B. It appears that guinea fowl were not kept in later periods.
The divergence in consumption of small vs. large domestic bovids between Harlaa and Harar/Ganda Harla  4 Identified fauna from Harlaa, Harar and Ganda Harla. Animal size classes follow Brain (1974) with small mammals (including small canid, small carnivore and small primate) equivalent to animal size class 1 (0-23 kg), small wild bovid, small bovid and medium mammal equivalent to animal size class 2 (23-84 kg) and large wild bovid, large bovid and large mammal equivalent to animal size class 3 (84-296 kg). No remains were recovered from animal size classes 4 (296-900 kg) or 5 (over 900 kg) - --86 7.4% 11 2.7% 43 5.4% 47 7.3% 103 13.8% 10 9.3% 5 3.8%    (Bekele & Yalden 2013). The hunted small bovids from all three sites favour drier open terrain with a combination of scrub environments, hillsides and grasslands. The non-bovid hunted animals (warthog, bushpig, aardvark, porcupine, hare, gennet, mongoose and leopard) range in habitat from open to scrub habitats. In general, none of the wild taxa recovered from these sites indicates the need for longdistance hunting trips as all share a common overlap in habitats of open and fairly dry grassland or hillsides.
Fish bones recovered from Harlaa (see Table 4) indicate exclusively the presence of oceanic/Red Sea/Gulf of Aden fish on the basis of those elements which could be identified to family or genus. Given the inland location of Harlaa, the fish remains recovered most likely arrived in the form of preserved fish (i.e. salted or dried). Fish remains from phases 1-4 of Area B contain entirely postcranial elements with cranial elements appearing only in phase 5 from this area. Fish remains from Areas E and F likewise contain only post-cranial elements. The appearance of some cranial elements in phase 5 of Area B may indicate a change in the preservation techniques of fish imported to Harlaa.

Herd Management Strategies
The age structure of domestic stock was calculated through assessments of both epiphyseal (post-cranial) fusion as well as from patterns of tooth eruption and wear. The numbers of neonatal (0-2 months) and infant/ juvenile (under c. 6-9 months) elements were recorded for each taxon as well as the numbers of fused and unfused proximal and distal epiphyses of each element recovered from Harlaa, Harar, and Ganda Harla. Elements with multiple fusion centres (e.g. radius) were in the main recovered only as incomplete bone fragments. For those elements where a complete unfused diaphysis was recovered, only the earlier-fusing epiphysis was recorded as providing the minimum age for the individual (Jones & Sadler2012a, b;Pummel 1987a, b;Zeder 2006).
From both a bulk assessment of fusion and age group survivorship it is apparent that in all contexts the majority of cattle survived into adulthood (c. 80-100%) ( Fig. 4; Supplementary Tables S1 to S7). No remains were recovered from neonatal or infant cattle and only a very small proportion were culled as juveniles (c. 0-5%). This pattern of high adult survivorship suggests that cattle at all three sites were either kept as animals of labour, status, or other secondary products. Age data on the basis of epiphyseal fusion indicates a different pattern in the management of goats (Capra hircus). Whereas cattle across all three sites and their phases were predominately retained into adulthood, a significant proportion of goats were slaughtered as subadults ( Fig. 5; Supplementary Tables S8 to S14). These data indicate a strong bias in the management of goats for their meat, in particular at Harlaa. This can be seen in the high number of sub-adult animal remains, particularly between 18 months and three years of age. At Harlaa c. 50% of goats were slaughtered prior to three years of age, with c. 15-30% slaughtered specifically within this 1.5-3 year age window. By contrast, fewer goats from Harar and Ganda Harlaa were slaughtered as subadults with c. 60-70% of animals surviving into adulthood. While fewer ageable remains were recovered from these sites compared with Harlaa, those present are sufficient to indicate a more cautious management profile with a greater interest in the perpetuation or increase of livestock numbers.
Too few remains of sheep were recovered from Harar or Ganda Harla to conduct a robust assessment of age. The distribution of sheep from Harlaa by age category (Fig. 6) demonstrates a greater focus in earlier contexts (Phase 2, Area B, Area E and Area F) on the retention of adult animals, while those sub-adult animals that were slaughtered were also kept until very late in their subadult development (Supplementary Tables S15 to S19). In samples from later phases (Phases 4 and 5 of Area B), by contrast, the management of sheep indicates a less cautious management with a greater focus on meat production. In these samples, c. 54-58% of sheep survive into adulthood with a slaughter profile more consistent with that seen from goats at Harlaa than for sheep from earlier contexts.
In comparison with survivorship profiles derived from fusion of post-cranial remains, age profiles were also FIGURE 4 Age groups of cattle based on post-cranial fusion data. The NISP of ageable elements for cattle in each sample is given in brackets.
FIGURE 5 Age groups of goats based on post-cranial fusion data. The NISP of ageable goat remains in each sample is given in brackets.
compiled for cattle, goats and sheep from Harlaa, Harar and Ganda Harla on the basis of dental eruption and wear. Age estimates from tooth wear were made using a variation of the quadratic crown height method (QCHM) modified for bovids by Gaastra (2016) as well as visual tooth-wear ageing (Grant 1982). This form of age profile provides a much narrower estimate of age-at-death for individuals in comparison to post-cranial fusion, as it determines age to within a few months. In keeping with standard practice, teeth have been placed into age groups following Payne (1973) and Halstead (1985). Comparison of cattle survivorship on the basis of dental eruption and wear (Fig. 7) demonstrates a slightly higher culling of sub-adult animals than was indicated by post-cranial fusion. While survivorship on the basis of fusion indicated c. 80-90% of cattle from Harlaa surviving beyond 40 months of age, from dental survivorship this is reduced to 50-83%. Most interestingly, dental survivorship indicates the slaughter of c. 11-17% (vs. 4.8% postcranial representation of this age group from Harlaa Area B, Phase 2 and absent in all other samples) of cattle within the first year of life from two early contexts at Harlaa (Area B, Phase 2 and Area E). These young animals are not found at Harlaa from the contemporaneous Area F sample, which shows a bias towards the retention of cattle beyond age three (83% compared with 66.6% from both Area B, phase 2 and Area E) compared with any other sample from Harlaa.
Age profile information obtained from dental eruption and wear for goats corresponds well with that seen from post-cranial fusion, while providing additional detail (Fig. 8). The bias towards the slaughter of sub-adult FIGURE 6 Age groups of sheep based on post-cranial fusion data. The NISP of ageable sheep remains in each sample is given in brackets.
FIGURE 7 Age distribution of cattle on the basis of dental eruption and wear. Teeth have been grouped into age brackets following Halstead (1985). The NISP of ageable remains from each sample is given in brackets.
animals seen from post-cranial fusion data is also apparent in data from dental eruption and wear, but also indicates that this bias is stronger than was apparent from post-cranial fusion data, with only c. 10-30% of animals surviving into adulthood at Harlaa compared with c. 50-100% from Harar and Ganda Harla. While not discernible from post-cranial fusion data, it can be seen from dental age data that there is a chronological difference in the exploitation of goats at Harlaa. Circa 30-50% of goats from Phases 4 and 5, Area B can be seen to survive into adulthood, but the adult proportion from earlier contexts at Harlaa (Area B, Phase 2, Area E and Area F) is less than 10% with only 3% surviving beyond four years of age in phase 2, Area B, and no animals above four years of age in the other two early samples. This suggests strongly that, while a proportion of goats at Harlaa survived into adulthood (up to c. 40 months), very few individuals survived long after this point. Too few mandibles and teeth from sheep were recovered from Harlaa (and none from either Harar or Ganda Harla) to allow for further information on the management of this taxon to be gleaned from analysis of dental eruption and wear.

Sex Proportions of Domesticates
No sexable pelvic elements of cattle were recovered from either Harar or Ganda Harla (Table 7), and those from Harlaa all come from adult animals (Grigson 1982). With the exception of Phase 2, Area B, male and female animals are equally represented amongst those animals which can be sexed. As with cattle, no sexable pelvic elements of goats were recovered from Harar or Ganda Harla. At FIGURE 8 Age distribution of goats on the basis of dental eruption and wear. Teeth have been grouped into age brackets following Payne (1973). The NISP of ageable remains from each sample is given in brackets.
Harlaa, both sub-adult (by fusion status and muscle attachment development) and adult pelvic elements were recovered which could be sexed. In all phases and areas of Harlaa, all recovered female pelvic elements came from young adult or adult animals. Pelvic bones from male goats came from both subadult and adult animals, though in the majority these come from subadult animals. These data, taken together with that from both dental and postcranial age distributions, suggests that extraneous male animals at Harlaa were preferentially culled as subadults with female animals left to survive to a slightly greater age. This is consistent with the age data indications of the management of goats for meat production.

Pathological Bone Modifications
Muscle attachment development, pathological alterations, and sub-pathological re-modelling were all recorded. Muscle attachment development could be determined for 13% of identified elements and was graded from light (indicative of subadult individuals) to heavy (indicative of adult or older adult individuals) on a scale of one to five. Pathological alterations to bones were infrequent (2 bones or 0.02% of NISP). The distribution of sub-pathological developments for metacarpals, metatarsals, first phalanges and second phalanges in response to the strain of pulling, or changes to foot bones in response to traction (Fig. 9) were scored following the criteria of the pathological index (PI) as detailed by Bartosiewicz et al. (1997;Bartosieicz 2008). Measurements of the distal ends of metacarpals and metatarsals were also taken where possible. Foot bones which produced a positive PI score for traction (2.5 or greater) were infrequent (Table 8 and 9). While the diagnostic foot bones for traction identification were present at both Harar and Ganda Harla, none scored a PI greater than 2.0 and no measurable distal metapodials were recovered from either site. There were no indications of the use of cattle for traction. At Harlaa, samples from Area B (phases 2, 4 and 5) and from Areas E and F all    (Table 8). However, the presence of these sub-pathological alterations does not permit distinguishing between the potential forms of traction which may have been in use. In general, the repeated practice of pulling excess weight behind the animal can be said to have caused the foot bone alterations present, and may have taken one or more forms, including pulling of carts, or ards or ploughs, or the circular pulling motion to rotate the stone for a stationary grinding apparatus.

Butchery
Cut marks are relatively infrequent at Harlaa (311 elements, or 3.9%) at Harar (11 elements, 3.3%), and Ganda Harla (1 element, 0.9%). The scarcity of butchery marks is unsurprising, as with the exception of butchery using heavy knives to sever and portion bones, cut marks produced during the processing of carcasses are generally infrequent as the aim of the butcher is to avoid direct contact with bones which will dull butchery tools (Seetah 2007). Therefore, the majority of butchery marks can be considered as 'accidental' indicators of the process of carcass skinning, disarticulation and defleshing. These marks relate to three stages in the processing of carcasses. The first stage is skinning and evisceration, in which the skin and internal organs of the animal are removed  10 and 11). The second stage is disarticulation of the carcass into either individual elements or a set of butchery units. The third stage is the removal of meat from bones. This is known as 'defleshing' or "filleting' and can occur either before or after cooking (Fig. 12). The identification of which stage in carcass processing is represented by a particular butchery mark (or marks) has been determined in reference to experimental butchery studies by Binford (1981), Nilssen (2000) and Seetah (2007). Butchery marks observed on animal ones from Harlaa, Harar and Ganda Harla was recorded as to the location, direction and type of butchering mark. These were categorized according to the three main stages of butchery. In comparisons of butchery practices between sites (and site phases) the most diagnostic stage was carcass disarticulation. An overview of disarticulation marks identified from Harlaa, Harar and Ganda Harla can be seen in Tables 10 to 13. A common pattern of segmenting carcasses can be seen. Following the removal of entrails and the skinning of carcasses, the head was removed primarily through slices to free it from the first cervical vertebra (atlas). This was in the majority accessed from the dorsal side of the animal for both small and large animals. Less commonly, and primarily with small bovids, the head was removed by chopping through the cervical vertebrae (Fig. 13). The anterior limb shows two areas of segmentation -from division of the shoulder through the removal of the humerus from the scapula or from the division of the carcass at the elbow by separating the distal humerus from the radius and ulna (Figs. 14 and 15). The hind limb indicates only the separation of this limb at the hip by the removal of the femur from the pelvis. In some cases, however, this appears to have been achieved by chopping the pelvic girdle through the pubis (in large and small bovids) or through the illial and ischial shafts to remove the acetabulum along with the posterior limb (Fig. 16).
Butchery was conducted using both small and large knives, with large knives appearing to be more common in the butchery of larger animals. However, the segmentation of the axial skeleton in both small bovids and large animals follows a mix of approaches and the use of both larger knives for chopping as well as smaller slicing methods. Moreover, skinning can be seen to take place at different portions of the foot regardless of the body size of the animal, with small bovids indicating skinning marks on the first or second phalange either from the anterior or posterior limb. This is also seen with large bovids and large mammals (equids and camels) albeit with a smaller sample size. The overall impression given by butchery marks is of a pattern of individualized rather than professional butchery, with different individuals having different methods and different choices in or access to butchery equipment.
Only one butchered element was recovered from Ganda Harla, a cattle calcaneus which was disarticulated by fine slices. 11 elements from Harar contexts had traces of butchery, of these 3 were traces of filleting (both small and large bovids). The remaining 8 elements were from both small and large bovids and predominately indicated disarticulation by fine slices with the exception of one sheep illial wing which was chopped through in order to remove the sacrum and one cattle illial wing which was likewise chopped through for the removal of the sacrum (Fig. 16). This pattern of disarticulation of the hindlimb and the removal of the sacrum (to divide the pelvis) was also seen at Harlaa, although alternative methods of carcass dismemberment were also indicated at Harlaa (chopping through the pelvis around the acetabulum, removal of the sacrum        with fine slices, chopping through the femoral capitus) which are not represented in the small sample of butchered bones recovered from either Ganda Harla or Harar.

Identifying Islamic Butchery Practices and Diet in the Faunal Assemblages
Halal butchery practices are specific only in the practice of slaughtering the animal, which is done via the severing of the throat to rapidly exsanguinate the animal (Francesca 2014;Kocturk 2002). This is achieved primarily through one of two approaches -a rapid and deep severing of the trachea, oesophagus and blood vessels with a knife from across the ventral surfaces of the throat (this being the primary technique) or the decapitation of the animal with a large implement (knife, cleaver, etc.) (Francesca 2014;Kocturk 2002). Of the 25 hyoid bones recovered from Harlaa (none were recovered from either Harar or Ganda Harla), eight have evidence of butchery. Of these, three bones bear cutmarks consistent with the slicing of the throat (one each from Area E, Area B, Phase 2, and Area B, Phase 5, Fig. 17). More specific guidance for the skinning or dismemberment of butchered animals is not given in the Qur'an or Hadith but can be reconstructed by comparison with other Islamic period faunal assemblages outside Ethiopia and the few proto-historic assemblages published from within Ethiopia. Analysis of butchery from late Aksumite and post-Aksumite phases of Bieta Guyorgis (Aksum) show some common traits with butchery evidence from Harlaa. The severing of the head and neck at the cranium-atlas junction using small slices can be seen at both sites. The severing of the elbow joint through fine slices to the proximal radius and the posterior margins of the distal humerus, as well as fine slices to the body of the hyoid bone from the removal of the tongue is also seen at both sites.
Several butchery practices evident in the Harlaa material are, however, not seen in the Bieta Guyorgis assemblage, but are found across multiple Islamic period sites in Arabia (e.g. Barbar), Anatolia (e.g. Gözlükule, Kaman-Kalehöuük), Mesopotamia (e.g. Tell Tuneinir), the Levant (e.g. Ḥorbat 'Ofrat, Shallale, Tall al-Fukkar,), and Iberia (e.g. Alartos, Núcleo Arqueológico da Rua dos Correeiros) (Bangsgaard 2001(Bangsgaard , 2015Estaca-Gómez et al 2019;Hongo 1997;Horwitz 2009;Loyet 1999;Moreno-Garcia & Gabriel 2001;Omar 2017). Generally, detailed discussions of butchery evidence are not common in zooarchaeological literature from Islamic periods (cf. Insoll 1999: 95-99). Regardless, there are sufficient comparative examples to allow for the identification of particular key common practices. These involve butchery through the use of heavy knives, including the severing of the elbow joint by chopping through the distal trochlea of the humerus (Fig. 15), severing of the shoulder joint by chopping through the neck of the scapula or the humeral capitus (Fig. 14), the severing of the pelvic girdle either by chopping off the femoral capitus or by chopping free the acetabulum (Fig. 16), chopping through the distal tibia (Fig. 18), and lateral sectioning of the axial skeleton either in the removal of ribs (Fig. 19) or for the division of loins and tenderloins from the vertebral column (Fig. 20). While it is difficult to conclusively determine that these forms of carcass dismemberment derive from cultural commonalities of the Islamic world given the infrequency of butchery discussions in zooarchaeological literature, the presence of so many common features across a range of sites which are all absent from Bieta Guyorgis suggests that the use of these butchery techniques are likely an identifiable part of Islamic cultural expansion into this region of Ethiopia. Although some of the butchery marks predate by several centuries the first evidence for Islam at Harlaa in the form of mosques or burials, the butchery techniques may have arisen initially with the movement of Muslim traders otherwise archaeologically unrecognised, and subsequently with the spread of Islamization to Harar, and, potentially, Ganda Harla, along with the adoption of new techniques with the spread of heavier butchery knives, or a combination of such factors.
It is noteworthy that Harlaa, Harar, and Ganda Harla share a predominance of ovicaprines, in particular goats, with various contemporary sites in the Islamic World (e.g. Bangsgaard 2015;Brown 2016;Hongo 1997;Horwitz 2009;Loyet 1999;Studer et al. 2013;Taxel et al. 2017;von den Driesch & Docker 2002). Similarly, in these comparative sites, goats were also mainly culled for their meat as older subadults while cattle were more commonly retained into adulthood, as were horses, donkeys and camels (Bangsgaard 2001(Bangsgaard , 2015Hongo 1997;Horwitz 2009;Loyet 1999). Dietary prohibitions of varying degrees of severity also exist within different sects of Islam (Francesca 2014;Kocturk 2002;Rodinson 2012;Wheeler 2014). These include, for example, those specifically outlawed by the Qur'an -carrion, blood and pork -as well as those forbidden by Hadith and jurists within individual sects. These include cats, dogs, marine mammals, domestic equids (donkeys, horses or mules), predatory birds, carnivorous mammals, reptiles and most insects (Francesca 2014;Kocturk 2002;Wheeler 2014). Some animals are allowed within different sects (e.g. marine mammals for must Sunnī sects), thus variations in the presence or absence of prohibited animals may be indicative of Islamic dietary practices, as well sectarian traditions. However, one complication with the use of faunal profiles for such ethnic or religious identification is that the observance of these rules is not necessarily universal across individuals, sites, or regions (Rodinson 2012 (Francesca 2014;Kocturk 2002;Wheeler 2014). However, pig remains are also systematically present across a great many sites of Islamic periods in north Africa, the Levant, Mesopotamia, Arabia and Anatolia, albeit at very low levels (e.g. Bangsgaard 2015;Brown 2016;Bouchnick 2018;Chaix & Studer 2001;Estaca-Gómez et al. 2019;García-Rivero et al. 2018;Hongo 1997;Horwitz 2009;Loyet 1999Loyet , 2004MacKinnon 2017;Studer et al. 2013;Taxel

Harlaa, Harar, and Ganda Harla in Comparative Regional Context
The faunal assemblages from Harlaa, Harar and Ganda Harla provide a window into animal economies in eastern Ethiopia during the 7th to 17th centuries AD. Given the scarcity of faunal studies done in the Horn of Africa from these periods there are few sites against which to compare the samples. At Mifsas Baḥri for example, a late Aksumite site in southern Tigray dated to between c. 550-700, the recovery of 34.6 kg of animal bones is referred to (Yule 2017: 76), but no further information on this material is provided. Similarly, excavations at Mezber have produced isotopic studies on the recovered remains of both humans and (studied) animals, although after more than a decade only the domestic chicken remains have been published (W0ldekiros & D' Andrea 2017). However, a wider regional comparison can be made between Harlaa, Harar and Ganda Harla and some other sites of late-and post-Aksumite northern Ethiopia (c. 6th to 9th/10th centuries AD). From comparison with multiple settlement contexts excavated at Bieta Guyorgis it can be seen that cattle were the predominant domesticate in late-and post-Aksumite periods, and sheep dominated slightly over goats (Cain 2000;Chaix 2013). Both of these patterns are in contrast to the findings from Harlaa. Cattle were the most common domesticate at both Harar and Ganda Harla, although the ovicaprines represented at these sites indicate a predominance of goats in common with Harlaa and unlike samples from Bieta Guyorgis. While fish remains were recovered from both Bieta Guyorgis (Cain 2000) and Harlaa, the fish species at the former are all local freshwater species, while those from Harlaa represented imported maritime species. Chickens are found on all sites, although the guinea fowl (Numida meleagris) is absent from Aksumite sites (Cain 2000;Chaix 2013;Woldekiros & D' Andrea 2017) and is only represented in the earlier occupational phases at Harlaa (up to the late 13th century AD).
Some differences in the management of domestic animals can also be seen. In both regions, cattle were predominately culled as adults, although the proportion of cattle culled prior to adulthood is slightly higher from Bieta Guyorgis samples (Cain 2000;Chaix 2013). From analyses of post-cranial fusion only c. 60-80% of cattle can be seen to survive into adulthood at Bieta Guyorgis sites compared with 80-100% of cattle surviving into adulthood at Harlaa, Harar and Ganda Harla. Sheep and goats from Bieta Guyorgis indicate a high culling of infant and juvenile individuals and young subadults (under two years of age) in contrast with the pattern of high culling of older subadult animals seen at Harlaa (Chaix 2013). This suggests that the management practices were different, although the lack of separated age profiles for sheep and goats from the Aksumite sites makes it difficult to confirm this.

Conclusions
The faunal assemblages from Harlaa, Harar, and Ganda Harla have provided significant information on animal economies and religious identity in medieval Ethiopia, and for the first time from an Islamic archaeological context. The appearance of a similar range of butchery techniques suggests that these were linked with the appearance of Muslim traders at Harlaa, and the subsequent spread of Islam to Harar and Ganda Harla. Intriguingly, at Harlaa, Islamic butchery techniques predate other markers of Islam such as mosques or burials by at least four centuries, suggesting that some Muslims might have been present much earlier, but are otherwise materially intangible. The presence of wild pig at Harlaa and Ganda Harla, tortoise rarely at Harlaa, but more commonly at Harar and Ganda Harla, and donkey at all three sites indicates dietary eclecticism, and at Harlaa and Ganda Harla the non-observance of Islamic dietary rules and/or the presence of mixed communities. At Harar, the absence of pig remains infers greater Islamic orthodoxy which would concur with its status as a city of Muslim scholarship and pilgrimage (Insoll & Zekaria 2019).
Social inferences may also be made. The absence of cross-site differences in animal body portions consumed at Harlaa areas B, E, and F, or at Harar and Ganda Harla suggests wealth or status may not have been a factor in access to meat. This correlates to a certain extent with distribution patterns of artifacts such as imported glazed pottery, beads, and shells at Harlaa, which do not show significant disparities across the site, but any such wealth correlations cannot be determined at Harar and Ganda Harlaa where these categories of prestige artifacts were generally absent (Insoll 2017). Similarities across the three sites were also evident in, for example, some of the species present, and in the hunting of small bovids.
The data also indicates differences among the sites, particularly between Harlaa, and Harar and Ganda Harla. For domesticates, goat dominates at Harlaa, and cattle at the latter two sites. All bird bones are absent from Ganda Harla and guinea fowl is only present at Harlaa, and only in the 11th-13th centuries. Imported, probably processed, marine fish are found at Harlaa, with no fish remains occurring at the other sites. Aspects of animal management strategies also appear to differ. The increased slaughter of sub-adult goats at Harlaa compared to Harar and Ganda Harla implies a more cautious approach at the latter sites where there was an emphasis on the increase or perpetuation of livestock numbers. Cattle appear to have only been used for traction at Harlaa. The patterns which are emerging from these faunal assemblages indicate the development of animal economies and dietary identities that are both connected with the wider Islamic world and rooted in the local environment. To explore these more fully it is imperative that comparable zooarchaeological analyses are undertaken at other late Aksumite, medieval, and Islamic sites in Ethiopia and elsewhere in the Horn of Africa.