Vocabularies of Enslavement & Unfreedom

By Elizabeth Urban

For the results I presented in my first and second essays, I used the search function on al-Maktaba al-Shamela to search for the key terms mawlāh and mawlayāt in specific texts. It is also possible to search these (and other) keywords across a particular genre of texts, or texts from a particular time period, as categorized by al-Maktaba al-Shamela. Therefore, I have begun playing around with al-Maktaba al-Shamela’s search function in order to gain some sense of how often specific terms for enslavement/unfreedom are used in different texts. For my third essay, I want to talk about some of the methods I have been using to try to understand the vocabularies of enslavement in early Arabic-Islamic sources, as well as some of the challenges I have been facing in this endeavor.

I must begin by noting that there are problems with al-Maktaba al-Shamela itself. The corpus available on al-Maktaba al-Shamela is vast, but it does have limits. For instance, I would have liked to search al-Iṣfahānī’s (d. 356/967) Kitāb al-Aghānī or Ibn Aʿtham al-Kūfī’s (early 3rd/9th c.) Futūḥ, but neither of these works is available on al-Maktaba al-Shamela. Additionally, the search function in al-Maktaba al-Shamela is known to return fewer results than other search engines.[1] However, I began with al-Maktaba al-Shamela because it is easy to use, readily available, and requires no extra knowledge of computer languages or coding.

There are also serious methodological issues that come along with doing any digital humanities work. One must design their study carefully, or risk simply “finding what they are looking for.” The computer will provide results for whatever terms are entered in the search, no matter how unrelated or unsuitable those terms may be. Conversely, the computer will not search for a term not entered in the search, even if it is the most salient term, making the results incomplete and misleading. Designing the search requires someone—or better yet, a team of someones—with a human brain and who is familiar with the sources.

That brings me to the specific searches I have tried, and the challenges I have faced. I began by searching the entire group of texts that al-Maktaba al-Shamela has marked as belonging to the first or second centuries AH. I searched for all the terms I could think of that refer to enslaved and/or freed persons, in both the singular and plural forms, in both definite and indefinite forms, in both masculine and feminine forms (where applicable), and with various prefixes and suffixes attached. My bank of search terms was as follows (only given here in the singular indefinite for brevity): ʿabd, ama, fatā(h), ghulām, jāriya, khaṣī, mamlūk(a), maʿtūq(a), mawlā(h), milk al-yamīn (or mā malakat al-aymān), muʿtaq(a), muwallad(a), raqīq(a), surriyya, umm walad, waṣīf(a). I welcome any suggestions for other terms to add to the list.

Immediately, my search was beset with problems. First, the most common term for male slave, ʿabd, also retrieves the hundreds of thousands of results for proper names such as ʿAbd Allāh (and ʿabīd yields results for ʿUbayd Allāh). It would be practically impossible to sift through each instance of ʿabd to remove all the ones referring to a proper name, though perhaps a programmer could create a shortcut. Similarly, in al-Maktaba al-Shamela’s search function, is treated the same as a tā marbūṭā, so that a search for the word ama (enslaved woman) also produces thousands of hits for umma (community) and ummuhu (his mother). Again, I trust that a clever programmer can write a code that treats a and a tā marbūṭā, but I am not that person.

Secondly, several of these terms for enslavement/unfreedom are ambiguous or polyvalent. Two very common terms for enslaved person, ghulām and jāriya, literally mean “boy” and “girl,” respectively. Even if a scholar carefully read each “hit” for these terms, it is often difficult to determine whether the person being described is an enslaved person or a youth (or both). And one of the challenges of studying early Islamic mawālī (or mawlayāt) is the sheer polyvalence of the term mawlā(h): it can mean cousin, ally, client, freed person, convert, or non-Arab Muslim.[2] As I mentioned in my previous essay, the term can also refer to either party in the manumitter-freedperson relationship or the patron-client relationship. While it’s usually easy to determine this one by reading the text itself, it means that one has to look through every single “hit” for the term mawlā(h) to understand if it is referring to a freed person or to some other meaning of the term. That is, one would hope the computer could do most of the labor of searching and tallying relevant terms, but this particular term involves a lot of human analysis as well.

How is one to overcome such difficulties? I tried one method on which I would love to hear my readers’ feedback. I created a search comparing only the plural definite terms of various words for unfreedom. It helps ameliorate some of the problems outlined above. My thinking was that if the sources use such terms in the definite plural form, they might be talking about a salient social group or category.[3] That is, it might give us some sense about how our authors imagined enslaved/unfree people as a group, rather than as individuals enmeshed in relationships of dependency. What the results indicated is that 69% of the hits referred to unfree/enslaved men, with only 31% referring to enslaved women. This finding suggests that we might want to reconsider the assumption that most enslaved people in early Islamic history were women.[4] When looking only at the subset of masculine terms, 27% of the hits referred to freedmen (al-mawālī, al-muʿtaqīn, and al-maʿtūqīn), while the remaining 73% referred to enslaved men per se. Conversely, for feminine terms, only 1% of the hits referred to freedwomen, while 99% referred to enslaved women per se. This finding suggests that either manumission was much less common for women than for men, or that enslaved women made a much greater impression on our authors than did freedwomen. (Perhaps freedwomen’s status was also subsumed under other statuses, such as wife or mother.) In any case, while my experimental workaround doesn’t overcome all difficulties, I wonder if these results might still be useful or at least suggestive of further avenues for research.


[1] See Maxim Romanov, “Never Trust al-Shamila Search Results,” The Digital Orientalist, December 1, 2018, https://digitalorientalist.com/2018/12/01/never-trust-al-shamila-search-results☝%EF%B8%8F/.

[2] Ibn Manẓūr lists manumitter/manumitted as the primary definition of “mawlā.” He also mentions that the term can mean friend or cousin. However, he does not mention “non-Arab Muslim” as one of his definitions, indicating that by the 7th/13th century this definition was no longer salient. Ibn Manẓūr, Muḥammad ibn Mukarram, Lisān al-ʿArab, 15 vols. (Beirut: Dār Ṣādir, 1955–1956), 15:408–10. For an extended analysis of the term mawlā and its possible meanings, see Elizabeth Urban, Conquered Populations in Early Islam (Edinburgh: Edinburgh University Press, 2020), 48–76.

[3] One potential problem with this methodology, especially as it pertains to Islamic law, is that Islamic jurists generally do not refer to cases involving entire social groups, but rather refer to specific, individual cases. It thus seems probable that legal sources will under-represent social groups, or at least will not use the kind of plural definite language suggested here as a potential methodology. I thank Intisar Rabb for bringing up this problem. I would respond by saying that we certainly cannot understand these plural definite terms to be representative of the actual size of social groups, but rather the relative size of social groups. That is, my working assumption is that the sources will equally under-represent all such plural definite social groups, at least giving us some sense of scale. Moreover, comparing the search results across texts from different genres can help us better understand the extent to which different authors working in different genres thought in terms of social groups vs. individuals.

[4] On the preponderance of women in global Medieval slavery, see Craig Perry, David Eltis, Stanley L. Engerman, and David Richardson, “Slavery in the Medieval Millennium,” in The Cambridge World History of Slavery, vol. 2: AD 500–AD 1420 (Cambridge: Cambridge University Press, 2021), 17. S.D. Goitein also noted that almost all of the Geniza records pertaining to slavery or manumission refer to enslaved women rather than men (“Slaves and Slavegirls in the Cairo Geniza,” Arabica 9, no. 1 (1962): 1­–20.

(Suggested Bluebook citation: Elizabeth Urban, Vocabularies of Enslavement & Unfreedom, Islamic Law Blog (Nov. 23, 2023), https://islamiclaw.blog/2023/11/23/vocabularies-of-enslavement-unfreedom/)

(Suggested Chicago citation: Elizabeth Urban, “Vocabularies of Enslavement & Unfreedom,” Islamic Law Blog, November 23, 2023, https://islamiclaw.blog/2023/11/23/vocabularies-of-enslavement-unfreedom/)

Leave a Reply