Arab Migration During Early Islam: The Seventh to Eighth Century AD from an Archaeological Perspective

1 History of Conquests and Migration

Historical sources narrate how the Prophet Muhammad and his early followers conquered important areas in Arabia during the 620s AD. After Muhammad’s death in 632 his generals, and later on the caliphs, continued the conquests and eventually occupied the whole Arabian Peninsula, Syria, Iraq, Egypt, Iran, and North Africa (e.g., Kennedy, 2007). Whenever a settlement was seized, so the sources tell us, its local leaders could choose whether to announce their surrender (sulḥan) and then to make a peace agreement with the conquerors (amān). Alternatively, when choosing not to surrender, settlements would be conquered by force (ʿanwātan) (Donner, 1981, p. 240; Kennedy, 2007, p. 19). As part of their peace agreement, some historical authors argue, communities agreed to be partly evacuated to enable settlement by the newcomers (Donner, 1981, p. 246). Lands owned by the church or the local authorities were confiscated and divided between the conquerors. Later on, additional lands were granted by the caliphs to their family members or could be purchased by other Muslims (Cahen, 1953, pp. 25–28; Donner, 1981, p. 241; Legendre, 2018).

The migration process of the conquerors is rarely described in the written sources (Donner, 1981, pp. 231, 245). For that reason, modern scholars interpret the notion of seventh-century emigration based on fragments of information and on their own common sense. Fred Donner, for example, uses texts which mention different salary levels as evidence for differentiating between early and later migrants. He argues that the warriors, most of nomad tribal origin, settled immediately after the conquests. Many more settlers, however, were immigrants who belonged to the same Arab tribes as the warriors and followed their relatives. The only question he thinks that the written sources cannot answer is the flow of this migration – either in several waves or as a steady thin stream (Donner, 1981, pp. 231–237). Kennedy emphasizes another interpretation: the conquests were conducted by an army of men and the tribal emigration occurred only later (Kennedy, 2001, p. 4).

The medieval authors provide scarce but valuable information. A twelfth-century historian, Ibn ʿAsākir, describes the army as accompanied by women and horses but excluding donkeys and sheep (Donner, 1981, p. 119). The text can be understood in two ways: as evidence for a direct migration of warriors and their wives, or, based on the absence of livestock, as evidence for a lack of settlement. The historian al-Ṭabarī (died 923) mentions anecdotes which coincide with Donner’s interpretation: the assemblage of tribes in the new settlement al-Baṣra in Iraq, as well as the dominance of salary-receivers in the new settlement al-Kūfa in Iraq, first, and of nomads, second (Donner, 1981, p. 232 and endnotes). Other scattered stories in the Arabic sources indicate migration but cannot answer the question about its chronology. The historian Ibn al-Kalbī (died 819) writes about one tribe which left the desert altogether and moved to al-Kūfa (ʿAthamina, 1987, 9, f. 28). Our historian al-Ṭabarī, mentioned above, notes the settlement of an Arabian tribe by the name ʿDra in Jordan (Hasson, 1984, p. 65).

One central type of occupation which modern scholars see as evidence for migration is the establishment of miṣrs, a term which often translates as “garrison town” (e.g., Kennedy, 2006, p. 23; Whitcomb, 2009). Modern historians suggest that the miṣrs were established in conquered areas in order to avoid assimilation of the newcomers with the native population and to prevent local rebellions (Crone, 1980, pp. 29–30, 1999; 2006, p. 11; Kennedy, 2001, p. 7). Importantly, some historians in the 1970s believed that these settlements were established for the nomad warriors and their families (Ashtor, 1976, pp. 18–19; Stern, 1970, p. 30). Later on, scholars explained the raison d’être of miṣrs as supplying a base from which militant campaigns could begin (ʿAthamina, 1987, p. 9; Bosworth, 1993). It is well accepted that different tribes settled within these settlements in defined territories, khiṭṭas (e.g., Akbar, 1989; Whitcomb, 2009, p. 248).

Some early sources use the Arabic words “emigration” (hijra) or “emigrant” (muhājir). However, the meaning of these terms is not as simple as one would wish. In its earliest appearance, emigrants are mentioned several times in the Quran. Regarding the date of the Quran, according to the Muslim tradition, it was revealed to Prophet Muhammad and was edited by the order of the third caliph, ʿUthmān, around the year 650. Some scholars support this date (e.g., von Sivers, 2021). The year 712 AD is the earliest date given for complete manuscripts (Grohmann, 1958, no. 17), making the end of the seventh century and surely the early eighth century a very likely terminus ante quem. In the Quran, muhājirs are first and foremost “believers” (Lindstedt, 2015, p. 70) who are in many cases mistreated and relate to a holy war (Crone, 1994, pp. 354–355). Some later Muslim traditions focus on holy war as the main meaning of the term, and in fact consider it a synonym to jihad, while others present additional messages. First, the word represents the physical move of the early believers from Mecca to Medina during the time of the Prophet. The counting of Islamic years starts from that presumed day and is thus named “of the Hijra.” Second, hijra means the physical migration to the miṣrs during early Islam. These emigrants, or their successors, enjoyed fiscal benefits and a higher social status than Arabs who chose not to migrate. Third, the term denotes conversion to Islam to some authors: the spiritual transformation from an infidel into a believer. Fourth, it can mean the physical flight from areas ruled by infidels (Crone, 1994. See also ʿAthamina, 1987; Hoyland, 2015, p. 102. Lane, 1893, p. 2879). Nonetheless, Lindstedt (2015) argues that in the seventh century, muhājirs was the word used by non-Arabs to define Muslim-Arabs.

Two additional types of information about migration during the seventh century derive from the narrative sources. One relates to a series of transfers – i.e., forced migrations – imposed on different communities. Some of the migrants were captives who were brought into Arabia, such as Sogdians from Iran (El-Ali, 1959, p. 253). Others were inhabitants of the Arabian Peninsula who were “transferred,” such as the Christians of Najrān (Donner, 1981, p. 236). A related type of information – which comes from additional texts – regards the flight of native populations from Syria and Palestine following the Arab conquests or the former unrest in the area (Keiko, 1991; Levy-Rubin, 2011, p. 168; Theodoropoulos, 2020).

Some geographical sources provide information about the ethnicity or tribal affiliation of different groups in various settlements. This may indicate immigration by Arabs but only at some point in history, not necessarily immediately after the conquests. For example, the geographer and historian al-Yaʿqūbī (died 897) comments how the inhabitants of the city al-Ramla in Palestine are Arab and non-Arab (al-ʿajam), and that the people under their protection (sing. dhimmi) are Samaritans (al-Yaʿqūbī, 1892, p. 328. See also Calasso, 2021; Cooperson, 2015; Reinfandt, 2020, pp. 231–232). He also lists various tribes from the Arabian Peninsula and the regions in Palestine where they settled (Hasson, 1984, p. 57 and footnotes 6, 7). The author similarly describes the Arab/non-Arab mixture in Iṣbahān in Iran and in cities in Iraq and North Africa. In al-Qayrawān, in today’s Tunisia, he specifies it being settled by Quraysh (the tribe of the prophet from Mecca) and others from Arabia (buṭūn al-ʿarab). In Zawīla, in today’s Libya, the inhabitants come from Khorasan and Iraq (al-Yaʿqūbī, 1892, pp. 345, 348. See Calasso, 2021, pp. 49–51).

A different method employed by historians as evidence for migration by Arabs or others, in Spain mainly, is the use of place names. The Arab sources describe the migration of berbers, the indigenous people of Northwest Africa (Berbers or Amazigh), to the Iberian Peninsula (for a critical overview, see Rouighi, 2011). Esquilache mentions how a Berber confederation of tribes gave a canal near Malila that is used today its name: Hawwara. A different example he discusses is names of settlements around Valencia. He claims that although the Arab and/or Berber immigrants gave Latin names to these places, not Arabic or Berber names, the words they used never referred to former hamlets or villae but to natural resources, suggesting the non-Roman origin of the settlements (Esquilache, 2021, pp. 145–148). Sitjes presents additional cases in Mallorca, which are related to the Manqūr group (Sitjes, 2021, pp. 33–36).

Finally, scholars consider information which relates to the Muslim faith or Arabic language as evidence for Arab newcomers. In central and northern Europe, the Christianization of the population in the early medieval period is identified through church structures and burials around churches (e.g., Andrén, 2013; Vargha, 2022). Similar methods are employed for Islam, using mosques and burials together with inscriptions of religious messages. Regarding mosques, some narratives tell us about constructing them within or instead of a local sanctuary (Guidetti, 2013; Magnusson, 2017). Another type of information regards the individual conversion of locals. Conversion is expressed mainly by the status of mawlā (client) – slaves or native men who converted to Islam and who sometimes reached high political positions (Crone, 1999; 2006, p. 15; Cooperson, 2015; Pipes, 1980; Wensinck & Crone, 1991). Even without the term mawlā, Arabic or Muslim private names outside of Syria and the Arabian Peninsula might imply their foreign origin or their conversion (e.g., Bulliet, 1979, pp. 65–69; Legendre, 2018, p. 405).

2 Historiographical Challenges in Writing the History of Arab Migration

The list of historical sources which imply migration in the seventh or eighth century is very rich. Unfortunately, this impression is somewhat misleading. Modern historians who study the first centuries of Islam face a significant challenge when using most of the above narrative texts: a long gap between alleged events and their earliest certain record in writing. Our earliest sources are ascribed to compilers from the late eighth and mainly from the mid-ninth century onward. Therefore, the texts as we know them were compiled at least 150 or 200 years after the events described in them. This means that even if stories in these texts have a “core of truth,” this truth would possibly have been modified during the two centuries with details and names added or reduced, with altered terminology, and with new ideological messages. Moreover, the writing was not necessarily intended to present reality to begin with but to deliver relevant messages to a contemporary audience (Borrut & Cobb, 2010; El-Hibri, 1999, pp. 216–217). Scholars demonstrate, for example, that the issue of surrender vs force had not been written about before the early eighth century, i.e., at least 70–100 years after the conquests. It seems thus likely that the texts represent legal concerns of landholders in the eighth century, perhaps the conquerors’ successors, not necessarily the reality of the conquerors themselves (Clarke, 2012, p. 42; Noth, 1973; Shoshan, 2016, pp. 6, 15).

While the earliest texts we possess were compiled two centuries after the events, modern scholars sometimes use even later sources. This choice is understandable: the later the text, the more detailed its descriptions. For example, the Muslim conquest of the Iberian Peninsula is recorded in Arabic texts as early as the ninth century. In these, the events are mentioned in a few lines or paragraphs. Therefore, modern scholars often consider a much more coherent text from the seventeenth century, which describes the events over 60 pages (e.g., Clarke, 2012, p. 24; Safran, 2021). Another example for elaborate descriptions relates to mosques. Mosques are already mentioned by eighth and early ninth^-century authors. A current study of mine shows, however, that details about their physicality, such as size, decoration elements, or construction materials, cannot be found in the earliest writing, but only starting in the mid or late ninth century. A little later, in the tenth century, these descriptions have become even more elaborate. In addition, authors from the mid-ninth century list many more toponyms with a mosque than earlier authors.

The increase in details by later Arab authors is explained differently by two “schools” of modern historians. These are, in very rough terms, the optimist and the skeptical (Clarke, 2012, pp. 1–2; Hoyland, 2015, p. 232. See also Donner, 1998, pp. 5–31; Humphreys, 1991, pp. 72–88). According to the first school, the late-medieval authors were better equipped with early sources than early medieval authors. In other words, their descriptions are authentic and true. In contrast, the skeptics – and myself among them – hold that the later the text is, the more it has been modified. With some exceptions, most of the supplemented details resulted from later imagination. Scholars explain the development of the texts in relation to specific contemporary social and political contexts (e.g., Clarke, 2012, pp. 40–44; Shoshan, 2016, p. 15). In other words, in most cases, the descriptions of late-medieval authors are less reliable than earlier ones. The main problem is that the historical ”facts” we learn, even the ones I presented earlier, often resulted from a mixture of early and late medieval sources.

The critical reading of historical narratives does not end with chronology. Another difficulty is the city-centric perspective of their authors (Clarke, 2012, pp. 40–44; Heidemann, 2009, p. 493; Legendre, 2018, p. 399; Nol, 2022a, p. 268). This bias distorts any attempt to reconstruct realities outside the metropolis. Therefore, settlement patterns, among them migrations and occupation, cannot be interpreted using narrative sources alone. One example comes from Egypt, where contemporary documents can be employed for comparison. The analogy shows that, contrary to information from the narratives, large estates could indeed be owned by Christian administrative elites, not only by Muslims and the caliph. It seems that Muslims became landlords in Egypt gradually, mainly from the early ninth century onwards (Legendre, 2018, pp. 404, 413).

To summarize, the earliest Arabic texts mention very few elements of migration in the seventh and eighth centuries. The Arab authors write about the establishment of new settlements (the miṣrs), about the density of nomads in these spaces, and about forced migration. These texts were compiled at least 150 years after the event and might not represent reality. The more common arguments of modern historians about migration – of tribes or of army families – are only implied in the texts.

3 Archaeological Texts

Unlike the narrative sources, which are late, we possess contemporary textual evidence from the seventh and early eighth century. This includes papyri and inscriptions – both of which I call “archaeological texts.” Relevant published papyri in Arabic come from Egypt. They were found in the refuse piles of several sites such as al-Fayyūm, Ahnās and Ushmūnayn (Sundelin, 2004, p. 15). Many of the earliest published texts in Arabic or in Greek relating to the Arab/Muslim rulers, were written by one governor in 709–714 AD, Qurra ibn Sharīk, and is presented below. Even earlier texts from the years 642/3 were published as well and are relevant to our topic.

Papyri written in Greek clearly mention the word “emigrants” in various contexts. The word they employ, muagaritun (Mωαγαριτῶν), or its variations, equals the Arabic muhājirūn (meaning muhājirs). The earliest references derive from two documents from 642 and 643 (Grohmann, 1932, in Crone, 1994, p. 359; Hoyland, 2015, p. 105; Trombley, 2013, pp. 27–28, 32–33). The later text (PERF 558) is a bilingual tax receipt for 65 sheep (see https://www.islamic-awareness.org/history/islam/papyri/perf558). While in the Arabic section, the author identifies the recipients as “we,” the Greek section calls the recipients once “Saracens” (σαραχηνῶν) and on another occasion “muagaritun and others who have arrived” (ἄλλοις ἀνερχομένοις). The earlier papyrus (PERF 564), written in Greek only, documents the receipt of 96 horses. In this case, the Arab recipients (at least five names are noted) identify themselves on three occasions as magaritai in the document.

Many more relevant texts are derived from the early eighth century. One context is the requirement to provide wheat, wine, and other foodstuff for muagaritun (Bell, 1912b, pp. 371–372, doc. 1433; 1928, pp. 7–8, doc. 1449. For the Greek texts see Bell, 1910). This food comes, in some cases, from the granaries of Babylon, in today’s Cairo (Bell, 1911b, p. 384, doc. 1404; 1912a, p. 132, doc. 1407). The texts clarify that the mentioned muagaritun had household slaves who were provided with allowances from the same sources (Bell, 1912b, p. 369, doc 1433; 1928, p. 5, doc. 1441). Another context is the assignment of muagaritun to fleets, side by side with sailors (Bell, 1928, p. 7, doc. 1449). Bell (1910, p. xxxiv) believes them to be land forces and not mariners themselves.

As discussed earlier, modern scholars interpret muhājirs differently: as believers, converts, emigrants to Medina or the miṣrs, or as warriors. In the mid-seventh century, the term papyri could mean “Muslims” but seems to mainly reflect military men (see Hoyland, 2015, p. 102). Some of the early eighth-century papyri similarly allow the interpretation as members of the military, but, different from the former, point to their households as well. Thus, the eighth-century papyri clearly reflect beneficiaries. This interpretation coincides with some of the Arabic narratives about muhājirs and the privileges they enjoyed from their status. It is important to note that providing the muagaritun with wine does not necessarily reflect on the (low) level of their “Muslimness.” In fact, we know very little about the consumption of alcohol during early Islam. We do know that it was prohibited at least from the ninth century on the one hand, and that it was consumed, on the other (Nol, 2022a, pp. 151–152; 2023b).

Another papyrus (no. 1447), which dates to 685–705 on the basis of the governor’s name, is of special interest. This is a long list of people entitled for an allowance, and their affiliations. The text lists women, mawlās and other individuals by name. In addition, it mentions “various muagaritun” (or muagaritun of various places), starting with the ones of Fusṭāṭ, in today’s Cairo. Other individuals are mentioned as coming from “Kor” (κορ), a possible abbreviation for “Quraysh.” Some are affiliated with “Anṣār” (Aανζαρ), a name for the first followers of the Prophet who moved with him to Medina (Bell, 1910, pp. 360–368).

This document indicates, first, that emigration took place from Arabia to Egypt before the year 705. In addition, women took part in this migration. The document implies, however, that in the eighth century, these individuals – certainly migrants – were not identified as muagaritun; this title was reserved for another group. This interpretation is supported by PERF 558 and the line about muagaritun and others. Papyrus 1447 additionally suggests that even people who had migrated from Mecca to Medina with the Prophet were not identified as muagaritun, as they had another title (Anṣār). All the above indicates that the term muhājirs cannot be understood as “emigrants” per se. The meaning of “warriors” – perhaps expanded in the late seventh century to include their households – seems more suitable. In other words, the term is not, in itself, evidence for Arab emigration. However, it may certainly be evidence for the Arab conquests.

The Egyptian papyri provide relevant evidence for the mobility of labor, as well. Sailors, for example, were sent from Egypt to other regions, sometimes as a tax payment (Fahmy, 1950, p. 97). In one papyrus, Qurra asks the addressee about the number of sailors who stayed in Africa (Aφρικη) and the ones that returned to their district (Bell, 1910, pp. 24–25; 1911a, p. 279, doc. 1350). Different documents discuss the expenses over shipments of craftsmen such as carpenters, sawyers, or iron workers to help building, among other projects, the mosques of Damascus in Syria and Jerusalem in Palestine (e.g., Bell, 1911b, p. 374, doc. 1366; 1912b, doc. 1433).

A second type of archaeological text is inscriptions, be they engraved, molded, or painted. “Formal” inscriptions in the seventh and eighth centuries, often created at the instruction of caliphs or governors (Ritter, 2016), mostly for dating purposes, will seldom be used in this study. Much more common in this period are “informal” inscriptions in Arabic (“graffiti”) on rocks and on architectural features. In many of these, individuals express their devotion to God and/or the Muslim faith. The relevance of this evidence to the question of Arab migration is discussed later with the introduction of physical remains of mosques and Muslim burials. Most of the graffiti are found in Arabia and the deserts of Israel/Palestine and Jordan (Hoyland, 1997; Lindstedt, 2022; Milwright, 2021). Another early source are tombstone inscriptions (epitaphs) which employ Quranic verses. The earliest examples have been found in Cyprus, Uzbekistan, and Jordan, dated as early as years 649, 676, and 690, respectively (Thesaurus d’Epigraphie Islamique, no. 1536, 17873, 53316, http://www.epigraphie-islamique.uliege.be. For later epitaphs from Egypt, see Halevi, 2007, pp. 14–25). This evidence implies the mobility of Arabs, the mobility of other Muslims, or conversion. At any rate, it provides very early evidence for the presence of the Muslim faith outside of Syria and the Arabian Peninsula. This indicates mobility which perhaps relates to the Arab conquests.

Additional inscriptions in Arabic from the early to mid-eighth century provide straightforward evidence for mobility from the Levant. Two texts have been documented on Aegean islands, Kos in Greece and Knidos in Turkey (Christides, 2006, pp. 54–55; Imbert, 2013; Thesaurus d’Epigraphie Islamique, no. 30132). The latter mentions the origin of the author from “the people of Palestine” (ahl Filasṭīn), circa 900 km removed in a straight line. Other inscriptions of the eighth century derive from Libya and Algeria (Fenwick, 2022, p. 211; Thesaurus d’Epigraphie Islamique, no. 29963, 42784). The Algerian text, found at Biskra, mentions the deceased coming from among “the people of Ḥimṣ,” Syria. The two Levantine toponyms are repeated in inscriptions of the same time from Saudi Arabia: in al-Aqraʾ and Jabal Ḥismā (Thesaurus d’Epigraphie Islamique, no. 54532, 56647). The distance between the approximate location of the toponyms and the find spot of the inscriptions is visualized on the map, as well as the direction of movement (Figure 2). The arrows on the map, however, are schematic and do not indicate specific routes of movement. Whereas these inscriptions imply mobility during the eighth century, they might also indicate an earlier settlement of migrants who retained the identity of a migrant community.

Figure 2

The approximate location of Ḥimṣ and Filasṭīn and find spots of eighth-century inscriptions which bear the two toponyms.

Both types of inscriptions – formal and informal – use the term masjid (mosque) from a very early stage. However, the term does not imply an adjacent structure for prayer and does not necessarily indicate the practice of prayer. In the Quran, the word can be understood as a sanctuary in general (Bloom, 2003). However, the distribution pattern of such inscriptions correlates with the pattern of physical mosques that relate to an excavation project, to be presented shortly, suggesting a possible correlation in their function as well. Inscriptions from the first three centuries of Islam use the terms masjid at 15 sites and miʾdhana (minaret) at 2 sites (Nol, 2023a). The two words are to be found from 698 to 816 (the first two centuries of Islam) in the western Arabian Peninsula and in Greater Syria (Figure 3). From the year 818 onward, masjid spreads also to North Africa and Spain, to Egypt, and to Pakistan (Figure 4). It is not possible to directly relate this evidence to migration. Nonetheless, if it does, it would mean Arabs started migrating from the Arabian Peninsula and Syria only in the ninth century.

Figure 3

Spatial distribution of inscriptions with relevant mosque terminology, late first and second centuries of Islam (698–816 AD).

Figure 4

Spatial distribution of inscriptions with relevant mosque terminology, third century of Islam (818–907 AD).

To summarize, the contemporary written sources clearly indicate Arab mobility into farther territories. The seventh-century evidence can be interpreted as related to military campaigns, if not to settlement. Inscriptions from the eighth century distinctly show movement of individuals from the Levant to North Africa and the Aegean, which could be additionally interpreted as evidence for migrant identity. Most evidently, the papyri point to an emigration of men and women from Arabia to Egypt before the year 705 AD. Nevertheless, contemporary texts provide us with only very specific and thin contexts, at least geographically.

4 Migration in Archaeology

The third source with which we can examine the question of Arab migration – and the main topic of this study – is archaeology. Unlike archaeological texts, archaeology presents a broader geographical scale. Unlike the narrative sources, it provides contemporary evidence of the events. Here I will introduce two sets of premises on migration in archaeology which I transform into two possible models. I will then present a number of case studies from AENAE which might prove to be related to migration and will compare them to the models. In short, the models have first and foremost been designed to assist in identifying Arab and Muslim migration in the archaeological case studies.

Archaeologists in various disciplines, including Islamic archaeology, tend to interpret changes in material culture of specific regions as the result of contacts with other communities (see, e.g., Brittain & Clack, 2022; Frieman, 2021, p. 28; Tolme, 2013). While in some fields, the contacts are perceived through economic lenses, for example, others tend to explain contacts as deriving from migration. Despite criticism from different directions over the years, along with shifting discourses in academia, it seems that some of the main notions and questions on migration in archaeology remain (Furholt, 2018). Two central approaches can be identified: the more traditional “cultural paradigm,” and a more critical approach of recent decades. The two approaches differ, first, in the phases they associate with migration. Second, they have developed different explanations for the roles of the place of “origin” as well as the “destination” of migration. As a result, the two approaches differ in their interpretation of specific material evidence or changes in it. Nevertheless, both approaches are optimistic in essence and hold that migration can be identified in the archaeological record.

For some clarification, I define “culture” as a composition of knowledge, meanings, and behaviors of groups in specific contexts which is often communicated through language, symbols, objects, and arts. According to the cultural paradigm, well-defined geographies, well-defined in time, are identified as “cultural areas” through a unique set of material culture. Migrants are, accordingly, well-defined cultural groups deriving from their own cultural areas and thus having their own unique material culture (see Aldred, 2021, p. 12; Anthony, 1990, p. 896; Frieman, 2021, pp. 124–126; Hakenbeck, 2008; Van Oyen, 2018). New material culture at a site therefore represents new migrants and the site is interpreted as a “destination.” If the same material culture is found at another site, preferably from an earlier period than at the destination, the second site is identified as “origin.” Aldred (2021, p. 13) adds that, very often, “a straight line is shown between a point of origin and a destination” with only variations in distances. In accordance with this line of thought, it is also argued that the migration process can be a fast event (Carvajal-López, 2021, p. 62).

Archaeologists then tend to interpret the arrival of new migrants as causing a cultural change at the destination. This interpretation makes the process of transition self-explanatory (Carvajal López, in press). One example, on which I will elaborate shortly, is the interpretation of irrigation methods diffused within Spain as caused by a migration of Arabs and Berbers.

Some of the criticism targeting the cultural paradigm has pointed out that migration, including the identities of the people involved, is interpreted as static. Another problem with the paradigm is its sole reliance on migration as a cause of innovation. Clearly, the diffusion of knowledge and ideology comes in various ways, not only through migration, e.g., with trade or other forms of exchange (e.g., Burmeister, 2000, p. 540; Hakenbeck, 2008, p. 17; Van Oyen, 2018). Even more so, cultural change occurs by different mechanisms. Whereas contact between groups can be one motor for an increasing flow of knowledge, some innovations are created independently, without contacts, while others are never adapted (Amati, Munson, Scholnick, & Habiba, 2019; Costin, Earle, Owen, & Russell, 1989; Erb-Satullo, 2020; Frieman, 2013, 2021; Jiménez Puertas & Carvajal López, 2011; Tolme, 2013; Van der Leeuw, 1981, pp. 281–300).

As one alternative for the cultural paradigm, scholars suggest employing an analogy with twentieth-century migrations, studied by the social sciences, along with historical migrations from the seventeenth to the nineteenth century (Anthony, 1990; Burmeister, 2000). Hakenbeck advocates this approach, explaining that “while we cannot simply transpose modern scenarios onto the in origin: pre-industrialised period as if they were universal, we equally cannot assume that migration in the past was limited to the movement of ethnic groups or demographic expansion” (Hakenbeck, 2008, p. 19).

The authors of the alternative theoretical framework suggest several aspects in their case studies which might be relevant to archaeology. One aspect is the identification of phases in the migration process. The first migration phase is either a planned expedition or a more spontaneous journey by “scouts.” The scouts do not necessarily settle for a long period in a destination. They could include merchants, craftsmen, missionaries, or warriors (Anthony, 1990, pp. 902–903; Burmeister, 2000, pp. 548–550). After the return of the scouts to their origin, or the use of another channel of communication, the second migration phase comprises settlers. The settlers emigrate directly to their planned destination, an act for which Anthony uses the term “leapfrogging model” (Anthony, 1990, p. 903). Archaeologically, the designated settlement of the second group can be identified by settlement clusters at the destination site which are located far away from the origin.

A second aspect in the alternative theoretical framework of migration is the demographic characteristics of the migrants in each phase. For example, young men are expected to act as scouts, the first phase of migration, whereas entire families could characterize long-term or permanent settlers, the second phase. This aspect can be detected in the archaeological record through the identification of burial demography (Anthony, 1990, p. 905; Burmeister, 2000, p. 550).

A third aspect is the links between origin and destination. The scouts, for example, keep a continuous relation with their place of origin, on their return or earlier. This relation allows for a flow of information from destination to origin which assists future migrants in their decision-making and in preparation. The links between origin and destination are also expressed in the migration of the next generations. These migrants follow the early settlers and move to the same destination as their predecessors, sometimes family relatives (Anthony, 1990, pp. 902–903). Archaeologically, this could mean a cluster of dwellings which comprise similar characteristics at a site. Another archaeological expression of the link between origin and destination could be a continuous flux of artifacts (Burmeister, 2000, p. 549). Barceló argues that communication networks necessitate infrastructure such as routes and ports (Barceló, 2001, p. 299), another easily detected feature in archaeology.

Migrants who returned to their place of origin could be observed through personal goods they imported from the destination they had just left (Anthony, 1990, p. 904). This interpretation stands in contrast to the notion that migrants take personal goods with them from the origin to the destination (Taxel, 2019, p. 233; Taxel & Fantalkin, 2011, p. 94). Burmeister raises another important part of these relations: the impact migration had on the origin area itself – economically, socio-politically, and culturally. On the one hand, leaving the region of origin might hurt the people and property which are left behind. On the other hand, returning migrants might introduce new ideas and new capital. Archaeologically, abandonment or decline of settlement sites might signify migration (Burmeister, 2000, pp. 545, 550). Therefore, changes in artifacts and technological innovations perhaps indicate places of origin, not destinations.

A fourth relevant aspect in the analogies with modern migrations is the cultural characters that migrants develop at the destination. According to Burmeister, there are two social scenarios. In one scenario, migrants adapt to the culture of the destination, at least in the public sphere. This can be seen, for example, in the facade architecture of public structures – as well as of private homes – which is identical to its surroundings (Burmeister, 2000, pp. 541–542; 2017, p. 61). The first scenario can be reflected archaeologically in the recognition of small groups within settlement sites through the interpretation of their private activities. In comparison to other activities at the site, these groups might have a unique set of vessels, unique foodstuff, and a different interior organization of dwellings. These finds should be similar to finds in the origin. According to the other scenario, migrants respond to the culture clash with the destination society by creating a new culture altogether (Burmeister, 2000, p. 546). This could be reflected archaeologically by a small group at the site with a material culture different from both origin and other groups at the destination.

5 Models of Migration in Archaeology

I have summarized the two dominant archaeological premises on migration into two models. The first one is based on the cultural paradigm while the second is based on the alternative theoretical framework. The models have been simplified and exaggerated. They have been designed to examine which of our case studies, if any, could be related to Arab or Muslim migration. The models visualize the place of change, the chronology of change, and the links and direction of influence between origin and destination. In both models, an emphasis is given to a single place of origin and a single destination. In order to offer measurable properties and maintain consistency, I randomly defined these origins and destinations as regions of about 500 km in diameter. This arbitrary definition will be tested and discussed along the article.

According to my simplified solution, the two main elements which should help to identify migration are specific objects or characteristics which can be observed in both origin and destination, and a drastic change in material culture in one of them. Principally, the first model highlights the link between origin and destination in two subsequent timeframes, and cultural changes at the destination. The other model points to continuous links between origin and destination and the cultural changes in the place of origin. As one could expect, periodization constitutes an important element in the identification of cultural changes.

The models present three axes: place (origin and destination), time, and material culture. The time axis is represented by “period x” (the moment of initial migration) in the first model, including its previous and following periods, and by phases (x.1 to x.3) in the second model. Material cultures are represented by “y” or “z,” where “y” is the material culture of the origin prior to migration and “z” is a newly emergent material culture in response to migration. Their combination with “a” or “b” denotes small changes over time. In some cases, two distinct types of material culture are expected to be found at the site simultaneously, as represented by the plus symbol (+).

The first model corresponds with the cultural paradigm which expects new material culture at a site to represent migrants. In this model (Figure 5), the migrants brought material culture “y” with them to the destination. Therefore, “y” can be found at the destination during period “x” and earlier at the origin (“before x”) but cannot be found previously at the destination. The possible variation in “ya” means that after the migrants left their place of origin, the remaining local material culture changed only slightly. The variation “yz” denotes the cultural change which occurred at the destination as a result of migration. Examples could be technological changes in the production of artifacts that were already produced in earlier periods, or conversion to another faith.

Figure 5

First model for the identification of migration in archaeology.

The second model follows the alternative theoretical framework and presents several phases of migration (Figure 6). Accordingly, no significant changes occurred in the destination culture during the first phase (x.1), when only “scouts” migrated. One type of material evidence could be infrastructure (e.g., routes) which enables communication channels between origin and destination. In contrast, the second phase, in which “settlers” migrate (x.2), sees several changes. During that time, most of the population at the destination displays culture “z.” However, small clusters at the site, which represent the migrant communities, have both cultures “ya” and “yb.” Culture “ya” existed at the place of origin in the previous period and was brought by the migrants to the destination. Culture “yb” exists simultaneously in origin and destination, indicating the regular links between the two. Returning migrants, who represent the third phase in the model (x.3) are expected to take culture “z” with them from destination to origin. In addition, they are expected to develop a new culture which combines both cultures of destination and origin (“yz”). The origin, which suffers from decline in phase x.2 due to the migration of some inhabitants, will enjoy wealth on their return in phase x.3. In cases when migrants leave the destination, the destination continues having only its own culture which has changed over time (“zb”).

Figure 6

Second model for the identification of migration in archaeology.

6 Archaeological Case Studies

The following section explores case studies from AENAE. The cases vary in their specific period, in region, and in type. They were selected due to former arguments about their relation to Arab migration, or for their potential as evidence based on their characteristics. The reader may notice the imbalance between the Levant or Spain and other areas in AENAE. Beside my better acquaintance with the former, the archaeological research of Islamic remains beyond Israel/Palestine, Jordan, and Spain is in its infancy. The topics comprise irrigation techniques, the architecture of “Desert Castles,” pottery and soapstone vessel production, and religious practices as reflected in burials and mosques.

I have excluded two types of evidence which could potentially be related to migration and/or technological change: glass and coins. Glass production in the Levant had its recipe and raw materials changed in the early ninth century, replacing natron (for alkali) and shells (for calcium) with plant ash, and sand (for silica) with quartz or chert (Henderson, 1985; Henderson, McLoughlin, & McPhail, 2004, p. 454). Whereas this is a significant innovation which should get more scholarly attention, its late date excludes it from our time frame. Regarding Arab coins, Gasc argues for a correlation between their spatial distribution in Spain and places where Arabs settled (Gasc, 2021). Fenwick (2022, p. 208) goes further and suggests that Arabic coins minted in North Africa imply the spread of Islam. However, the minting of coins rather represents administrative operations and coin distribution provides information about economic networks (e.g., Nol, 2022b). Neither, I believe, is conducive towards identifying human agency. In fact, Arabic coins were frequently used in northern Europe during the medieval period among Christian and pagan societies (e.g., Knutson & Ellis, 2021, pp. 9–10).

The relevance of each of the selected case studies to migration is not always obvious and needs examination. This is conducted through a systematic set of categories offered by the theoretical models above. Dates are essential for establishing a timeline for a site and for detecting possible changes in it, such as an eighth-century change which would imply Arab migration at this point in time. For this reason, I place an emphasis on dating the finds. The dates I consider are based on archaeological dating tools, not on chronicles or on analogies with other sources. As a proxy, I consider the Arabian Peninsula and the Levant as places of origin for Arab or Muslim migration during the seventh and eighth centuries. As I will show, most of the cases proved to have too little secure evidence to support questions of migration. Nevertheless, some enable the migration interpretation and at least one case study supplies clear evidence for migration.

6.1 Irrigation Techniques

Archaeological research in AENAE has investigated various techniques for fetching and using water both inside and outside settlements. Some of these techniques are assumed to relate to the diffusion of knowledge or social change. Many techniques, such as wells, proved so successful that they have been employed for millennia with only very nuanced changes (Wilson, 2008). Some of the other water-fetching techniques are more limited chronologically and may assist, at least in theory, in solving questions regarding social transformations. This potential, however, is not always met due to several challenges. One is that for a very long time in most of the region, any sophisticated technology has been considered “Roman” (generally the second century BC to the fourth century AD) or ”Late Roman” (fourth to seventh century AD) so that precise dating was not pursued (Fenwick, 2013, p. 10). Another challenge is that even when features are dated, the results are often too broad for any significant typology. One of the solutions, dating water systems according to an analogy with relevant texts (e.g., Gorzalczany, 2011, p. 211; Kirchner, Virgili, & Puy, 2021), prevents the independent and empirical archaeological research we are aiming at. In addition to these challenges, secondary literature often lacks an overview and synthesis of water installations at different sites or in various periods.

One branch of scholars who advocate migration as the motor for cultural change is formed by archaeologists of Islamic periods in Spain. A central argument relates to irrigation in southern Spain as, first, it differs from previously employed dry farming, and second, is linked to Muslims and/or the spread of Islam. Thus, although irrigation solutions were known previous to Islam, it is an accepted assumption that they became widespread during the Islamic period. Their popularity was caused, archaeologists believe, by a migration of Arabs and Berbers (Barceló, 2001, pp. 306–307; Carvajal-López, 2023, pp. 33–34; Esquilache, 2021, p. 131; Kirchner, 2021, p. 12). Beside the chronological factor, scholars show that in northern Spain, a region under Christian authority, irrigation was not used (Jiménez Puertas & Carvajal López, 2011). Similarly, it is claimed that the Christian conquest in the thirteenth century interrupted the “normal evolution” of one such irrigated area “by the destruction of the Andalusi society that had built it” (Esquilache, 2021, p. 149).

This approach suffers from two main flaws. First is the invalid dichotomy between irrigation and dry farming. Dry farming consists of different agricultural practices and methods of water manipulation, and medieval farmers in southern Spain therefore had to use a variety of strategies, often in parallel. The irrigated areas were located close to the house while other farming methods were used in farther fields and orchards (Retamero, 2021). Superficial irrigation supplemented by rainfall in semi-arid and Mediterranean zones provides more security since watering the field before sowing results in a better yield (Ersten, 2012. See examples from modern Israel in Nol, 2022a, pp. 186–187). Above all, choosing irrigation solutions must fit a series of requirements and “forms of social management associated with different forms of irrigation,” such as a labor calendar (Retamero, 2021, p. 176). The motivation for irrigation might therefore be related to risk management and exchange systems, to demographic needs, or to taxation, among others (see Retamero, 2021, pp. 188–189). These issues then reduce the position of irrigation in the hierarchy of water management solutions.

The second flaw relates to the imprecise dating of the distribution of irrigation methods in Spain between the eighth and twelfth century – unknown both archaeologically and historically (e.g., Guinot Rodríguez, 2021, p. 119). Contemporary documents supply only terminus ante quem dates, for example of the 1150s for Tortosa or the 1240s for Manacor (Kirchner et al., 2021; Sitjes, 2021). Arguing for an eighth-century process is thus uncertain at best. The lack of dating makes the comparison to our models very difficult. Even more so, since irrigation was known in Spain before the eighth century (e.g., Barceló, 2001, pp. 306–307), the models would not be able to detect the change. Finally, research has not yet provided links with another region which could be identified as either origin or destination. In short, irrigation in Spain cannot be examined through our models.

A different case study, which may be more suitable for our examination, regards qanat irrigation in the Levant. The method (originally qanāt, sing. qanā, a channel in Arabic) includes the following elements. The method includes the following elements (Figure 7): (1) a “mother well” which is dug into the groundwater levels, (2) an underground channel that transfers water from the mother well to an open canal in a slight slope, (3) a series of vertical maintenance shafts dug along the channel, and (4) an open canal which allows water to flow from the channel outlet to the fields (Ahmadi, Nazari Samani, & Malekian, 2010; Avner, 2016; Avni, 2018; El Faiz & Ruf, 2010; Lightfoot, 1997; Porath, 2016).

Figure 7

The qanat system (drawing: Ron Haran after English, 1968).

The origin of the method had been ascribed to Achaemenid Persia in the sixth century BC, and its other appearances have been interpreted as a diffusion of knowledge from there (Charbonnier & Hopper, 2018; Ersten & Morgan, 2021). However, excavations of relevant remains show that the technology developed in different areas and at different times. The question of diffusion vs parallel inventions brings forward different arguments. Some scholars claim that variations in the different cases, and in their construction techniques, precludes it being a single method (Charbonnier & Hopper, 2018; Ersten & Morgan, 2021). Avni (2018) argues, in contrast, that the typology of various qanats is strongly related to chronology. Therefore, he suggests that much of the expansion of the qanats relates to specific authorities in specific periods – the Umayyads in the seventh to early eighth century.

Qanats have been known in the Levant from Israel/Palestine and Jordan (Figure 8). Their dating is based on the date of excavated or surveyed settlements in their environs. This suggests that the technology entered the Levant either in the Roman period before the fourth century AD (Driessen & Abudanah, 2018, p. 148), during the seventh or early eighth century AD (Avner, 2016, p. 24; Avni, 2018; 2020, p. 117; Porath, 2016), or during the late eighth or early ninth century in the “Abbasid period” (Nol, 2014, 2015). Along with the second scenario, the diffusion of the method in the region has been assigned to the Arab conquests (Avni, 2018). This, however, does not yet resolve the question about the agents of diffusion: Arab migrants or the Islamic “state” (Nol, 2014).

Figure 8

Qanat shafts near Ma’an (APAAME_20070419_FFR-0136, photography: Francescar Radcliffe, courtesy of APAAME).

In order to test if the qanat technique could ever fit our models, we shall refine some of the suggested interpretations. We would assume that the technique arrived in the Levant (a) in the eighth century and (b) from Iran or areas under Persian influence such as the Arabian Peninsula. This would agree with the first model, in which the migrants brought culture “y” with them to the destination, where it was unknown before. In contrast, the second model necessitates a continuous link between the two regions, which cannot be detected.

6.2 Desert Castles and Courtyard Structures

The second case study brings forward an architectural form which has been suggested to derive from the Arabian Peninsula and to spread in the Levant: the so-called “Umayyad Palaces” or “Desert Castles.” These are monumental structures which consist of a square plan, external towers or buttresses, and at least one courtyard (Figure 9). The title has been given to sites in modern Jordan, Israel/Palestine, Syria, and Iraq, including the most famous examples of Khirbat al-Mafjar, Jabal Says, and Qaṣr al-Ḥayr al-Sharqī. The possible functions that have been suggested for them include palatial residences, agricultural estates, trade stations, a safe space for Muslims, or early urban regional centers (Nol, 2022a, pp. 22–23; Walmsley, 2007, p. 99).

Figure 9

Schematic representation of various types of “Desert Castles” (drawing: Ron Haran). (a) Jabal Says (after Genequand, 2006, Figure 1:1), (b) al-Qaʿ (after Whitcomb, 1996, Figure 2:P), (c) Khān al-Zabīb (after Genequand, 2009, Figure 21: northern structure), (d) Umm al-Walīd (after ibid, Figure 10), and (e) Qaṣr al-Mshattā (after Genequand, 2006, Figure 5b:2).

The structures have been dated to the early eighth century and are often attributed to a single ruling family, the Umayyads. Inscriptions are the most reliable dating tool in these contexts – indeed providing dates from the 720s – but these are extremely scarce. Most of the sites certainly continued to be in use into the next centuries and evolved during later periods, as the few excavations have shown (Genequand, 2012, p. 99; Whitcomb, 2016). Although the architectural similarities between many of the Desert Castles imply patronage networks, it is generally agreed that they were not homogeneous artistically or functionally and perhaps should not be defined as a single group (e.g., Genequand, 2006; Hillenbrand, 1981, p. 65; Milwright, 2010, pp. 35–36; Walmsley, 2007, p. 105). Despite various modern terminologies and later dates, these structures resemble others in Uzbekistan from the ninth or tenth century, if not a bit earlier (Field & Prostov, 1948, Figure 5; Mirzaakhmedov, 2012) and in the Arabian Peninsula (e.g., Figure 9b), dated to the ninth to tenth century, based on a single excavation (O’Kane, 2014; Whitcomb, 1996. See also Avni, 2014, pp. 202–203; Northedge, 2009). A smaller and simpler variation contains rooms around a courtyard, such as in Maʿan or Ḥumayma in Jordan, which Genequand (2006) entitled quṣūr (sing. qaṣr, castle or palace) (Figure 9d) and which can also be found at a number of early Islamic sites in southern Israel (Nol, 2014).

A similar style of architecture could be identified at sites of the third to fourth century in Saudi Arabia (Northedge, 2009), one of them being Qaryat al-Faw. The excavated site includes structures which have been interpreted as a market, a palace, and a temple. The possible market is a fortified structure with seven towers, a gate, an internal courtyard structure, and a cistern. The internal rooms are interpreted as shops and resemble rooms in the Desert Castles. The site, published only partially, could be dated generally from the second century BC to the fifth century AD, based on partially published pottery and ¹⁴C results. Coins from the site date to the first to early fourth century AD. A foundation inscription in ancient South Arabian script from the temple was dated to the first to fourth century AD whereas the architecture and sculptures have parallels from the first century AD (al-Ansary, 1981, pp. 28–29). Al-Ansary thus advocates the migration scenario, suggesting that Muslims who planned their later cities such as al-Kūfa or al-Fusṭāṭ “were inspired by their own local heritage which they carried with them from the Arabian Peninsula” (ibid, p. 22). Northedge supports this general chronological sequence albeit with different nuances. He believes that al-Faw and similar Arabian sites did not invent the fortified structure but imitated Roman forts on the Arabian frontier. The forts, he argues, were then further developed in the Levant by the Umayyad family during the seventh century (Northedge, 2009, pp. 248–249). In other words, the idea of these forts was migrated either by migrants or by another mechanism initiated by the rulers.

The Desert Castles might roughly fit the chronological scheme of our first model. In that case, the origin would be the Arabian Peninsula and the destination in the seventh century includes the Levant to Iraq. Culture “y” was exported and later developed into culture ”yz” with the smaller square-shaped structures. One problem would be that their distribution falls into a region much larger than my pre-defined 500 km. The definition of a single destination is random, however, and therefore cannot pose a strong obstacle for interpreting the Desert Castles as related to migration. The main problem is their stylistic variety as well as their resemblance to architecture in other areas which question their suitability as a single type of material culture to begin with.

6.3 Portable Artifacts: Ceramics and Soapstone

Burmeister suggests that aspects of function and technology in material culture, “free of symbolic meaning,” do not change frequently and thus may represent migration best (Burmeister, 2000, p. 553). The following section is devoted to the production techniques of ceramics (in Spain) and to the introduction of soapstone (into the Levant and Egypt). While studies on the two sets of material are limited as yet, the former relates to technological innovation and the latter to the mobility of artifacts.

In southern Spain, pottery production had changed before the tenth century in a number of ways. The innovations included new types of oil lamps, a portable cylindrical oven, tableware, and new decoration techniques in color. All these characteristics and objects are also known from Egypt or the Levant (Carvajal-López, 2021; Salinas, in press). In theory, one could argue for their link to migrants, but many more interpretations are also valid (e.g., a market-related one), especially when a more precise date for the changes is absent.

One study, from seven sites in the region of Granada, offers better dating for eighth and ninth century pottery. It demonstrates three modifications: a change in the clay recipe in the ninth century, an alteration of big storage jars, and developments in olla cooking pots (Carvajal-López, 2021). Figure 10 presents changes in the form of ollas from the fifth to the eleventh century AD. In the first phase, before the year 800, the volume of the container remained the same, along with its globular profile and even its main production technique on the wheel. However, the height of its neck grew and allowed for a spout, a morphological change which mirrors changes in cooking or in serving. Moreover, the later olla types seem to have thinner walls, which might indicate a higher wheel speed. Another characteristic, not shown in the figure, is a new technique for finishing the pot bases – scraping – in addition to the traditional technique of a wire cut. These shifts become clearer in the second phase, 801–925 AD. No other changes could be detected in the second phase (Carvajal-López, in press). It is noteworthy that not all developments could be observed at the Granada sites in parallel (idem, 2021, p. 63; in press). These changes come in contrast to a technological stability in the earlier period, between the fifth and eighth century. Carvajal-López connects them with the arrival of migrants to the region (Carvajal-López, 2021; in press). Our models, however, necessitate more data to relate the changes to farther regions and the aspect of bringing the knowledge (“y” or “z”) from elsewhere.

Figure 10

Changes in the typology of olla cooking pots, Spain (drawings by José Carvajal-López, Miguel Jiménez Puertas, and Angel González Escudero, with permission).

Soapstone vessels make a different case. These objects generally include greyish/greenish softstone types such as steatite, chlorite, or talc (Phillips & Simpson, 2018). The most common artifacts for the eighth to ninth century are “pots” or “bowls,” “incense burners,” and open “lamps” with a triangular or a star form. The incense burner is a small container with four legs and, usually, has a handle (Figure 11). Identical shapes were found in clay objects (Le Maguer, 2011). At the Late-Antique site Qaryat al-Faw, identical vessels were made from limestone, some engraved with South Arabian inscriptions (Al-Ansari, 1981, pp. 72–73). Braziers and lamps were recovered at key medieval sites, including Qaṣr al-Ḥayr al-Sharqī and Fusṭāṭ. In contrast, soapstone bowls are widespread among various sites, with many examples from southern Israel as well as al-Ṭūr in Sinai, Egypt (e.g., Fabian & Gilʿad, 2010; Kawatoko, 2007, pp. 10–11; Nol, 2015, pp. 53–55).

Figure 11

Soapstone “incense burner” from Ramla, Israel (photography: Clara Amit, courtesy of the Israel Antiquity Authority).

It is generally accepted that the provenances of the soapstone were south-east Arabia and south-east Iran, areas where such vessels had already existed before the eighth century and continued in production and use long after the tenth century (Simpson, 2018, p. 180. See also Le Maguer, 2014; Walmsley, 2000, p. 331). However, petrographic analyses of vessels from southern Israel imply a more local source, such as Sinai or Southern Jordan (Israel, Nahlieli, & Ben-Michael, 1995, p. 10). The common use of the soapstone bowls in southern Israel/Palestine supports the idea of local production, as well (Nol, 2022a, p. 209).

The locality of bowl production in Israel/Palestine, its clear former use, and the limited region it was found in make it a great candidate for our models. Following the first model, people might have migrated from the Arabian Peninsula to the neighboring desert of Israel/Palestine. They would have brought with them their preference for pots made of soapstone and the techniques for producing these vessels (“y”), exploiting a more local source and creating a “yz” culture. Alternatively, and pursuant to the second model, people might have migrated primarily from southern Israel/Palestine to the Arabian Peninsula. Only with their return to Israel/Palestine would they have brought with them Arabian knowledge and taste (“z”). Due to the known technology in the Arabian Peninsula prior to Islam, however, the first model gains more support.

6.4 Religious Practices: Mosques and Muslim Burials

One way to detect cultural and social changes in a region is through the religious practices of its inhabitants, such as funeral rituals and prayer. One of my premises was that the earliest Muslims lived in the Arabian Peninsula and Syria, the “origin.” Thus, one of the possible proxies for migration of Arabs during the seventh and eighth centuries would be the spread of Muslim rituals in this period. Above, I have explored the use of the term masjid in inscriptions from the eighth and ninth centuries. Other religious texts, such as the use of Quran verses, were not considered here. In this section, the first type of evidence examined is mosques related to an excavation project. The second one is Muslim burials. The relation of both to Islam, or the reasons for their identification as such, are discussed below.

Mosques can be identified archaeologically based on elements that are known from later structures: a qibla (a wall oriented to Mecca) and a miḥrāb (a round or square niche in the qibla). In some cases, the structures comprise a ṣaḥn (an open courtyard) and a riwāq (a portico around the courtyard). Another common method archaeologists employ is that of identifying a late-medieval mosque or a cathedral and excavating earlier structures below it which might have been mosques too – if not other sanctuaries (Nol, 2023a). The structures considered for analysis here all date from the seventh to the ninth century based on independent dating tools provided by their excavators (Nol, 2021). The dating tools include datable inscriptions, ¹⁴C results, coins, pottery, or other identifiable portable artifacts, and index fossils of artistic motifs or construction techniques. The structures are mapped according to centuries. Mosques with a terminus post quem or a broad pottery-based date are marked differently from the ones with more certain dates.

The earliest three mosques are dated to the seventh century but are without certain dates: Aqaba in southern Jordan, Nahal La’ana in southern Israel, and al-Yamāmah in Saudi Arabia. I thus have included them with eighth-century mosques. The distribution map of the seventh to eighth century (Figure 12) shows a distinct cluster of mosques in Greater Syria, with 23 mosque sites. The map highlights, in addition, two structures in the Arabian Peninsula and three mosques with only terminus post quem dates in Iran, Uzbekistan, and Morocco. In short, by the end of the eighth century, mosques had possibly spread within the Arabian Peninsula and into Syria. A different picture can be drawn for the ninth century, when mosques spread to present-day Spain, North Africa, Central Asia, and possibly to east Africa, with a total of 43 mosque sites (Figure 13). In other words, Islam spread broadly, but later than our timeframe of the seventh to eighth century.

Figure 12

Spatial distribution of datable mosques of the eighth century AD.

Figure 13

Spatial distribution of datable mosques of the ninth century AD.

These two patterns resemble the ones for masjid inscriptions (above): a local distribution during the eighth century and a widespread distribution during the ninth century. The conclusion for masjids regarding migration repeats, as well: if the pattern represents Arab migration beyond Syria and the Arabian Peninsula, then the process took place in the ninth century. The results for the eighth century do not enable us to use the models on migration since the whole distribution area – Syria and the Arabian Peninsula – is considered to be the “origin.” If migration occurred, it happened within the space which I have identified as originally Arab. The spread in the ninth century is beyond our research frame. Even when considering it, however, it shows a broad distribution which answers a great number of continuous destinations. This type of distribution does not suit the models, which are designed to answer a single destination.

A second type of evidence for our question is Muslim burials, a central method for identifying individuals who adhered to the Muslim faith. In the Quran, burials are mentioned indirectly in three verses (Petersen, 2013, p. 242). The earliest relevant texts were written or compiled by ninth and tenth century authors such as Ibn al-Shāfʿī (died in 820), Ibn Saʿd (died in 845), and al-Ṭabarī. The Arab authors express legal norms on funerary customs in their time, including the construction of low graves, constructing them with a niche in the qibla wall, and burying corpses lying on their right and facing Mecca (Halevy, 2007, pp. 91, 188–189, 321). Another source, some suggest authentically earlier, is the poet al-Farazdaq (died ca. 732) who mentioned the Kaʿba (in Mecca) “toward which every grave is turned” (Gorzalczany, 2007, p. 71, followed by Petersen, 2013).

The archaeological evidence from different periods largely supports the ideal of corpses facing Mecca, while the other aspects are more diverse (Simpson, 1995). Skeletons facing Mecca thus can serve as a proxy for locating Muslim burials, at least from the eighth or ninth century onward. After the identification of such burials in early Islamic contexts, our discussion about migration requires the dating of these burials. One case study is a cemetery in Tashbulak, south-eastern Uzbekistan, with approximately 400 graves. Out of 40 skeletons that were examined, three infants and one adult had their face straight up, and 33 were oriented close to the direction of Mecca (220–300°; Mecca is oriented 242°). The deceased came from both sexes and varied in age. Six of the adults were dated with AMS. The earliest date lay between 680 and 770 (with 95.4% accuracy) and the other five between 760 and 970 AD. The authors suggest them as evidence for an early conversion of the local population (Bullion, Maksudov, Henry, Merkle, & Frachetti, 2022).

A second case study comes from Nimes, France, where 20 medieval to modern burials were found. Three of the burials were identified as Muslim due to the skeleton position: positioned on their right side and facing Mecca. In two of the pits, a niche had been dug into the wall. Samples that were taken from the three burials gave similar ¹⁴C dates between 680 and 760 with 95.4% accuracy. The three skeletons were of males. According to some of their characteristics, the authors suggest their African origin and thus their identification as migrants (Gleize et al., 2016). In Pamplona, Spain, a cemetery was found above the Roman bathhouse. In the cemetery, 190 graves were excavated, showing a similar position of skeletons facing Mecca. One skeleton of a male exhibiting signs of violence was dated with ¹⁴C to 650 to 770 (Faro Carballa, García-Barberena Unzu, & Unzu Urmeneta, 2007). Finally, in Tell Qarassa in southern Syria, the skeletons of a young male and a young female were excavated from later contexts of a Neolithic cemetery. The two skeletons face Mecca. They were both identically dated with ¹⁴C to between 666 and 766 AD, suggesting a close time of death (Srigyan et al., 2022).

All four case studies present a very similar pattern of Muslim burials from the late seventh century at the earliest to the mid-eighth century latest. Their distribution, although sparse, is very broad (Figure 14). Despite the lack of physical evidence for such burials in the Arabian Peninsula, for the purpose of testing the models, I could presume an “origin” based on the historical knowledge of early Islam originating in Arabia, or the “origin” I have defined earlier of the Arabian Peninsula and the Levant. Another challenge which arises is the limited number of burials excavated in most of AENAE. This does not allow for determining whether there are additional Muslim burials between Uzbekistan and the Levant, for example, and thus how broad the distribution was. Still, we might assume that the current burial sites are the main ones in existence and then interpret them as separate destinations. Within these ideal parameters, both models are possible. According to the first model, Arab migrants brought culture “y” (Islam) with them from the origin (Arabia or Arabian Peninsula and the Levant) to the various destinations during the seventh or eighth century. According to the second model, returning migrants brought culture “z” (Islam) with them from an Islamic destination to their various origins. In the case of Uzbekistan, if the skeletons belong to Uzbeks, they represent local converts, as the excavators have suggested. Locals could have converted to Islam either influenced by Muslim/Arab migrants, or by their migration to an Islamic destination. Importantly, both models overlook the interpretation of converts from elsewhere, as suggested for the African origin of the individuals in France.

Figure 14

Spatial distribution of Muslim burials dated to 650/680 to 770 AD.

7 Arab Migration Through the Archaeological Records: Some Conclusions

Our archaeological inquiries into the AENAE of the seventh to ninth century have provided a variety of case studies for new forms of material culture and evidence for the practice of a new confession. These comprised irrigation solutions and pottery in Spain, the qanat irrigation technique, the so-called Desert Castles, and soapstone bowls in the Levant, and mosques and burials in various regions. One of the purposes of this article was to examine if any of the changes could be interpreted in relation to migration. Another purpose was to test the validity and usefulness of two theoretical models I offered to the question of Arab migration. The two main elements I defined as significant for identifying migration were objects or characteristics which can be observed in both the region of origin and of destination, and a change in material culture in one of them. The first model was based on the cultural premise. It highlights the link between origin and destination in two phases and cultural changes at the destination. The other model, based on more recent “alternative” premises on migration, points to continuous links between origin and destination and the cultural changes at the origin.

The archaeological data from early Islamic AENAE proved to be insufficient in most cases. Out of seven selected case studies, three were ineligible for the models and another three required interpretive refinements or assumptions. Data which did not qualify for our models include irrigation solutions in Spain which are not securely dated and, in any case, preceded Islam; pottery in Granada which lacks links to sites elsewhere; mosques which circulate first only in the region of origin and have too broad a distribution later. The qanats presented difficulties with dating, with the assumption of having a single earlier provenance, and with proving to be one group, but otherwise would answer the first model. The Desert Castles were similarly difficult to interpret as a single group and their distribution is broader than my arbitrary definition, but otherwise they answer the first model, as well. Regarding Muslim burials, the highly limited data provide no information about an earlier provenance or their real possible distribution. Otherwise, they could answer both models. The only case study which showed sufficient data for our models were the soapstone vessels, suggesting a link between the Arabian Peninsula and southern Israel/Palestine. These links coincide with both models, but the earlier provenance in the Arabian Peninsula favors the first model.

One result which this research indicates, surprisingly for me, is the insignificant role that early Islamic mosques play in understanding migration, conquests, or innovation. Our data present the relatively late spread of both physical mosques and the term masjid outside the Arabian Peninsula and the levant, only in the ninth century. The earlier spread of Muslim burials, before 770 AD, supplies a much better proxy for the early spread of Islam. Fenwick indeed suggests that mosques were built as a political message by rulers, mostly “in urban settings,” and thus provide only a partial representation of society (Fenwick, 2022, pp. 210, 214).

This research indicates that employing the models necessitates precise dating and a sufficient coverage of broad regions. Without precise dating, some of the data could answer both models – which provide two very different historical interpretations. Whereas the first model is supported by the observation of an earlier provenance – often based on wide ranges of dates, the second model relies on continuous links between regions, which are difficult to detect in the archaeological record. The soapstone case study demonstrates short-distance links, but broader distributions within limited regions – e.g., mosques in the eighth century – are difficult to detect. Moreover, the size definition for origin/destination regions was random, highlighting random aspects of the models. Clearly, as has been argued by Aldred (2021, p. 13) and demonstrated to some extent on burials in France of possible African Muslims, the models do not accommodate chains of origins or destinations. They are designed to highlight a direct migration from origin to destination or the return-migration from destination to origin. Moreover, the models cannot assist in “proving” migration over other interpretations for the mobility of objects.

I believe that in order to better test the two theoretical models, we should shift the focus of our studies from “innovations” to possible links between sites of different distances. Such links would include any material culture which different sites share. Links between sites would refine our interpretation of past migrations or object mobility to distant destinations as well as to closer ones. Another direction for future archaeological research on migration would be further use of ancient DNA for tracing mobility and social grouping. In AENAE such studies have already been conducted (Bycroft et al., 2019; Srigyan et al., 2022).

Whereas the question about Arab migration cannot be answered yet by archaeology alone, it is possible to consult other early Islamic sources as well as studies of modern migration. Egyptian papyri confirm the migration of men and women from the Arabian Peninsula to Egypt during the seventh century. Likewise, eighth-century Arabic inscriptions suggest either an eighth-century migration or an earlier one from the Levant to the Aegean islands, Libya, and the Arabian Peninsula. Both the narrative sources and the papyri discuss the Arab conquests. Arab warriors can be interpreted as the “scouts” of the migrant communities in different places, as modern migrations suggest. Another characteristic of Arab migration which is noted in the narrative sources is settlement in tribal clusters (khiṭṭas). While this narrative is not confirmed by other sources, it is supported by analogies to modern migrations, where members of migrant communities settle in proximity to one another. In short, our sources indicate the migration of Arabs in the seventh and eighth centuries. Hopefully, new methods in the archaeology of AENAE could provide an independent type of evidence for this question in the near future.