Computational methods readily lend themselves to concrete applications in certain sub-fields of study: e.g. analysis of the fluctuation of commodity prices or demographic change. * A major issue arises, however, when it comes to enacting computational procedures to abstract objects of inquiry that constitute much of what historians spend their time and energy on, such as “schools of thought,” “movements,” “civilizations,” etc.1 The challenge appears to be two-fold:

1. Capturing (aspects of) the highly complex sets of interrelations that make notional entites like “movements” so compelling as objects of study;
2. Conveying the observed dynamism graphically (rather than in so many words).

Recent reflection on these issues led to an investigation that my colleague Uzay Cetin (Bilgi University, Istanbul) and I are currently preparing for publication, in which we use computational methods to quantify aspects of the abstract quality of “leadership” among early Iraqi Sufis (i.e., from the 3rd/9th-4th/10th centuries). Initial results of this analysis were presented at the 2018 MESA conference in San Antonio, TX, and you can view the slides here.

This post builds on that presentation in two parts:

Melchert’s “transition” thesis: A positivist historical model

Thucydides conceives his history under the model of the medical history […] within the frame of a natural and teleologically governed progress, passing inexorably from a beginning through a middle to an end.
- Joel Fineman, “The History of the Anecdote: Fiction and Fiction,” in Harold A. Veeser (ed.), The New Historicism (London: Routledge, 1989), p. 55.

Melchert offers a model of historical change that is not dissimilar to that of Thucydides; the basic supposition is that the most prominent early Sufis serve as able proxies for analysis of the underlying machinations of the more nebulous development of the notional entity of “early Sufism.” In the article, Melchert connects three arguments relating to leading early Sufis, drawn from anecdotal material: (1) that a “transition” between “ascetic” and “mystical” discourse can be located in these figures’ recorded sayings at a specific point in time; (2) that “conflict” between ascetics and mystics crested with the Sufi “inquisition” (Ar. miḥnah) of 264/878, prosecuted by Ghulām Khalīl (d. Baghdad, ca. 274/888); and (3) that the Sufi Abū al-Qāsim al-Junayd al-Baghdādī (d. Baghdad, 298/910-11) promoted a “reconciliation” between the two camps by articulating a program for “Classical Sufism” in a “mystical” idiom that was virtually unintelligible even to his immediate predecessors. The first of these contentions concerns us most here, although, as will be discussed, it is not entirely separable from the others.

Melchert begins discussion of his “transition” model by asserting that:

[A] transition from Islamic asceticism to Islamic mysticism has now become a scholarly commonplace. (p. 51)

He situates his contribution to this framework by claiming that, before the middle of the third/ninth century, “the extant record” relating to notable 3rd-/9th-century Muslims is “overwhelmingly ascetical,” after which time it becomes “clearly mystical” (p. 57), as in the sayings attributed to Dhū al-Nūn al-Miṣrī (d. Egypt, ca. 245/860). Melchert’s confidence in this claim rests on a distinctly positivist epistemology. For example, he takes for granted the historicity of sayings attributed to figures in Sufi prosopographical works - mainly al-Sulamī’s (d. Nishapur, 412/1021) Ṭabaqāt al-ṣūfīyah and Abū Nuʿaym al-Iṣfahānī’s (d. Isfahan, 430/1038) Ḥilyat al-awliyāʾ. 3 Furthemore, he asserts that “ascetical” or “mystical” elements of these sayings occur in “different proportions that can be measured” (p. 52) or “weighed” (p. 54), thus reflecting an “ascetical” or “mystical” outlook for a given figure.

The terms “ascetic” and “mystical” deserve some explication here, as they are central to the machinations of the proposed historical model. Melchert’s notion of an enduring conceptual “split” between “asceticism” and “mysticism” – which is clearly legible in the concepts of “transition,” “conflict,” and “reconciliation” – is drawn from the ideal types used in the works of Max Weber (d. 1920), namely “Religious Rejections of the World and their Motivations” ( Essays in Sociology ), and Economy and Society. Melchert contends that Weber’s ideal types in fact facilitate the analysis of the data relating to early Sufi activity; recall that he posits that figures from the first half of the third/ninth century were “overwhelmingly” ascetic, while those from the latter half were “clearly” mystical.

With these facts in mind, we can now comment more broadly on Melchert’s model. In essence, Melchert’s conception of a “transition” is that of a dramatic re-organization of Islamic piety, akin to the “phase transitions” that are illustrated in the graphic below. According to Melchert’s model, and using language that has been applied to phase transitions, with the passage of time we should first observe a steady state, here referring to a stabilized and overwhelmingly ascetical vocabulary, followed by (1) degredation of the steady state, and (2) the re-appearance of a steady state, here referring to a stabilized and clearly mystical vocabulary.4 This type of transition thus roughly resembles the topological models in the second (middle) column of the graphic.

Figure 1 – Types of Observed Phase Transitions
Phase Transitions
Source: F.W. De Sousa Lima, “Phase Transitions in Equilibrium and Non-Equilibrium Models on Some Topologies,” Entropy 18/3 (2016). (link)

An empirical critique of Melchert’s “transition” model

Through recourse to the idiom of “transition,” a concept developed in the physical sciences, Melchert gestures toward the complex interactions of early Sufis. While it is probably true that all models are “teleological” in the sense described by Fineman, above, the teleology of Melchert’s “transition” model is driven largely by the rhetorical trope of an “opposition” beween asceticism and mysticism, specifically the neat decomposition of the dynamics of early Sufi activity into the beginning (= transition), middle (= conflict), and end (= reconciliation). Perhaps because of the appeal of this rhetoric, Melchert’s “transition” model has become central to many subsequent overviews of Sufi history, in spite of persistent pushback. To my mind, previous critiques of Melchert’s model have adequately called attention only to one of the two most serious issues with the positivisim at its heart, each of which affects the viability of the “transition” thesis.

The most vigorous criticisms of Melchert’s “transition” model highlight his problematic postulation of the relation of the Weberian conceptual framework to the structure of the hypothesied “transition.” Recall that Melchert insists that elements of an ascetic or mystical outlook or “phase” can be measured / weighed, and that this computational procedure is commensurate with the Weberian paradigm. However, the empirical separability of the ideal types of “ascetic” and “mystic” was never enforced by Weber in his historical studies. It is on this exact point which Bernd Radtke has caustically rebuked Melchert’s “transition” model; recently, Christopher Adair-Toteff has persuasively reiterated the point.5 Likewise Sara Sviri has come down harshly on Melchert’s model, despite the fact that she, like Melchert, insists that “asceticism” and “mysticism” took different “tracks” within the Islamic context. Instead, she rejects Melchert’s positivist approach, and declares that, “a historical and paradigmatical ‘transition’ from asceticism to mysticism seems to [her] a”fallacy.”6

As persuasive as I find each of these critiques to be, neither of them addresses a second – and to my mind more fundamental – issue with Melchert’s model: the weakness of the empirical evidence that he adduces in support of it. Given the scope of his claims, one would expect Melchert to furnish evidence that is (a) contemporaneous with the phenomenon being studied, (b) abundant, and (c) systematically collected in order to support the very specific contentions related to his “transition” thesis. Unfortunately, what he does provide in this article satisfies none of these conditions.

(a) With respect to the sources of data used for this analysis, Melchert posits that “the extant record” of early Sufi activity may be “conveniently take[n]” (p. 53) from al-Sulamī’s Ṭabaqāt, Abū Nuʿaym’s Ḥilyah, and Abū Naṣr al-Sarrāj al-Ṭūsī’s (d. Ṭūs, 390/980) Kitāb al-Lumaʿ fī al-Taṣawwuf, which were composed either in the late fourth/tenth century, or the first third of the fifth/eleventh century. This rationale can no longer be defended considering the wealth of published primary sources on Sufism that predate the aforementioned works.

(b) Recognition of the existence of early sources not utilized by Melchert leads to the related issue of his positivist exploitation of later works. In particular, the quantity of the anecdotal data that he extracts from the later sources is not sufficient to support significant aspects of the model. In fact, he infers the “transition” from an analysis of the “ascetial” and “mystical” elements of only 43 (!) sayings. These sayings are chosen from only a dozen figures, and for no figure does Melchert’s analysis exceed the “measuring” or “weighing” of four sayings, even though there are frequently dozens of recorded sayings for any one figure in al-Sarrāj’s Kitāb al-Lumaʿ, al-Sulamī’s Ṭabaqāt, or Abū Nuʿaym’s Ḥilyah (see below).

(c) Moreover, the method that Melchert uses to extract quotations from these sources is also fraught from the positivist standpoint. His method for choosing sayings for analysis appears to be relatively ad hoc, the seeming result of a search for sayings that best accord with the hypothesized Weberian “phases.” This operation is especially evident in the example of one figure who died at almost precisely the mid-point of the third/ninth century: al-Muḥāsibī (d. Baghdad, 243/857), in whom Melchert identifies a “change” (p. 55) in the proportion of ascetic to mystical sayings ( nMelchert = 4; cf. nṬabaqāt = 24; nḤilyah = 87 [!] ).7
The issue of the influence of the Weberian framework on Melchert’s sampling of data again arises when one notices that he only considers sayings that occur in the notice (Ar. tarjamah) for a given figure, rather than sayings that originated with someone else but which that figure in turn transmitted, as documented in isnād s. This feature of Melchert’s approach likely affects his analysis of figures like al-Junayd, mentioned above, who is identified as a major tradent for the sayings of earlier figures like al-Muḥāsibī and al-Sarī al-Saqaṭī (d. Baghdad, 249/863) – both of whom Melchert identifies with at least tacit ascetic leanings. Therefore, the failure to include in his analysis “unoriginal” sayings likely makes al-Junayd seem more “mystical” than would otherwise be the case. Although this scenario is admittedly speculative, it is clear that Melchert’s account is guilty of the same charge. In summary, and contrary to his own presentation of his method, the “measurement” or “weighing” of the data at Melchert’s disposal was not conducted in any systematic or reproducible way, a fact which should call his conclusions into question as a matter of course.

In this section, I have highlighted a number of relevant criticisms of Melchert’s method in convening “transition” thesis. First is his adoption of a Weberian notion of the opposition of “asceticism” to “mysticism” as the basis of an empirical model, even though Weber himself did not subscribe to the use of this opposition in historical studies. Following this, Melchert’s choice of data with which to test this model is based primarily on “convenience,” rather than an exhaustive search for relevant data; the existence of earlier Sufi literature likewise complicates Melchert’s presumption that the later literature provides an accurate a depiction of the situation of third-/ninth-century Sufis and their affiliates. Finally, his anecdotal approach to and interpretation of the data extracted from these sources is a red flag, and appears to be guided in no small part by the choice of the Weberian paradigm. At each step in the analysis, Melchert’s methodological choices decrease the explanatory value of his model of the dynamics of early Sufi activity.

Nevertheless, there is an important caveat that must be appended to the preceding critique. Although Melchert cannot be said to have proven his case - much less persuasively argued it…

there is still a possibility that “the scholarly commonplace” of a “transition” between ascetic and mystic modes of religiosity occurred in the way Melchert describes.

To his credit, Melchert has posed his hypothesis in terms that are easily testable, assuming that a more representative dataset can be convened. In order to evaluate the validity of his “transition” model, the following steps need to be taken:

1. Compile a dataset from early Sufi sources that offers a more robust basis of analysis with respect to both (a) the total number of data points considered, and (b) relative temporal distance from Sufis of the third/ninth century.
2. Test Melchert’s contention about the effects of “ascetic” or “mystical” discourse in the development of Sufism in a more systematic fashion by rigorously quantifying these aspects of the lexicons of individual figures.
3. Evaluate the goodness-of-fit between the new dataset and the transition model, described above.

In the next section I briefly treat the first two topics, as they are dealt with at greater length in the forthcoming article. However, the last point will be the subject of a fuller elaboration.

Testing Melchert’s model:
Compiling data for a metric of the “consistency” of Sufis’ sayings

The dataset used to test Melchert’s “transition” thesis comprises eleven isnād-based works: this number includes al-Sulamī’s Ṭabaqāt, as well as his Dhikr al-Niswah al-Mutaʿabbidāt al-Ṣūfīyāt, in addition to nine other works that were composed before or contemporaneously with these titles. Cumulatively, this dataset comprises 2,044 isnād s, featuring approximately 2,600 historical figures8; the number of people in an isnād ranges from two to twelve. In order to avoid duplicating Melchert’s selection bias, we propose a parameter for identifying a “leading” Sufi that does not depend on the presence of a notice / tarjamah in al-Sulamī’s Ṭabaqāt; this incurs the added benefit of considering women as potential Sufi leaders. Specifically, the parameter we used requires that a leader is recorded – considering all of the the isnād s in which he or she appears – as passing information to at least five separate “followers”; this operation is more fully explicated in our paper. This restriction returns a subset of 92 individuals from the original dataset, whose dates of death have been recorded; these data can be accessed here.

Based on the interest in testing Melchert’s specific thesis about the “transition” from asceticism to mysticism, we conceptualized a metric that was suitable for evaluating his claims. Given Melchert’s insistence on the possibility of “weighing” the “proportion” of ascetic and mystical sayings in the sources, we decided to term this metric consistency. Deriving this metric from our data involved three steps.
First, we created a list of terms that plausibly fit Melchert’s notion of “ascetic” or “mystical.” Drawing on seminal works in the field, such as Paul Nwyia’s Exégèse Coranique et langage mystique (1970) 9, we compiled a list of lexemes with which to conduct the analysis – thirty ascetic and thirty-one mystical. Some lexemes proved more productive than others: for instance, the mystical root ḥ-q-q appears in many more forms than does the ascetic root h-j-d. These data can be accessed here.
As mentioned, we wanted to use this list to “measure” or “weigh” the consistency of the ascetic and mystical content of the leaders’ sayings in a more systematic fashion than Melchert accomplished. In order to do this, we took the second step of tokenizing the texts of the sayings attached to the isnād s using MADAMIRA. 10 We then ran the tokenized texts through a lemmatization protocol using Python, and evaluated matches in the texts of the sayings against the aforementioned list; again, this process is explained at greater length in the paper.
Finally, the scores derived from this computation have been “normalized,” meaning: a figure who uses only “ascetic” lexemes, to the exclusion of “mystical” lexemes – so identified in the list – evinces a score of “1” (“maximally consistent - ascetic”); a figure who uses only “mystical” lexemes to the exclusion of “ascetic” lexemes – so identified in the list – evinces a score of “-1” (“maximally consistent - mystical”); a figure who uses “ascetic” and “mystical” lexemes with equal frequency evinces a score of “0” (“maximally inconsistent”). A schematic representation of this computation is provided below:

Figure 2 – Schematic of tokenized “ascetic” and “mystical” sayings
tokenized-ascetic-mystic
- Green dots represent leaders
- “a” indicates a lexeme tokenized as ascetic
- “m” indicates a lexeme tokenized as mystical

If the prediction of Melchert’s model is correct, then we should see the following dynamics reflected in a visualiation:

1. Around the beginning of the 3rd/9th century, “leading Sufis” should show scores of consistecy above 0.5 (“overwhelmingly” ascetic).
2. At the middle of the century, “leading Sufis” should cluster at or near the “maximally inconsistent” score of 0.
3. By the decade of the 870s – i.e., the time of Ghulām Khalīl’s miḥnah, numerous “leading Sufis” should show score of consistency around -0.5 or greater (“clearly mystical”).


Dynamic visualization of Sufi leaders’ “consistency” using gganimate

In thinking about the best way to visualize the consistency metric, the obvious candidate seemed to begganimate, written by Thomas Pedersen (package site), mostly because its syntax is largely similar to that of the widely-used ggplot2 package. A number of accessible tutorials have been written for gganimate, but the inspiration for this particular visualization was written by Pedersen himself.

In creating the visualiation first step is to load the packges (NB there are apparently some issues installing through Windows), along with the data, and check that it is loaded correctly. The column marked as het, the computationally-derived score for “consistency” described above, contains the dependent variable:

library(gganimate)
library(ggplot2)
library(readr)

leaders <- read_csv("mesaData_NONA.csv")
head(leaders)
tail(leaders)
leaders$het

Consistency animation using gganimate

The code for creating the dynamic visualization with gganimate is fairly straightforward, assuming some prior knowledge of ggplot2. However, there are a few quirks to this particular visualization, which I will briefly explain after the code. Most of the specifications / aesthetics for this visualization are taken directly from Pedersen’s example, and I found that they looked very nice out of the box. The minimal theme, in particular, worked well for the presentation of this data, but other combinations could help, depending on the type and density of the data being presented, I suppose.

ggplot(data = leaders, aes(x = death, y = het, group = Sulamī)) +
  geom_line(aes(color = Sulamī, size = followers)) +
  geom_segment(aes(xend = 1000, yend = 0.0), linetype = 2, colour = 'grey') + 
  geom_hline(yintercept = 0.0) +     
  geom_point(size = 2) +
  transition_reveal(Sulamī, death) + 
  coord_cartesian(clip = 'off') +
  theme_minimal() + 
  theme(plot.margin = margin(5.5, 40, 5.5, 5.5))+
  labs(title = "'Consistency'of Sufi leaders' sayings through time",
       x = "Year of death (C.E.)",
       y = "(+ mystical)    Consistency    (+ ascetic)",
       colour = "Tarjama in al-Sulamī's TS?",
       size = "Leader's # of followers"
  )

geom_line(color = Sulamī) : the variable Sulamī in this dataset is logical, and denotes whether or not the “leader” has his or her own notice / tarjamah in a work by al-Sulamī. Since the varible is logical, the result will be two different colors; the default color scheme in ggplot2 is a kind of Miami Dolphins-ish orange and blue, which isn’t terrible but could be better.

geom_hline : this is to provide a visual reference to the value 0; aesthetically, it should function to make it easier for the audience to find the “tipping point” of “maximum inconsistency,” which Melchert identifies with al-Muḥāsibī.

transition_reveal : this is important for the animation process; a / the variable that indicates the passage of time in the dataset - here death - needs to be specified in this call. This function makes sure the correct layers or types of data will remain in the visualization as the images are rendered in an animation.

The previous code produces the following dynamic visualization:

Figure 3 - “Consistency” of Iraqi Sufi leaders’ sayings over time
sufi-consistency

This graphic charts approximately 350 years’ worth of data in what we termed the “consistency” of the sayings for those whom early Iraqi Sufis identified as their leaders. I was really excited to share this graphic at MESA because it seemed to embody the qualities of a good visualization outlined by John Thiebault: transparency, accuracy, and information richness. That is, it seems to tick all the boxes with respect to the question of: Did Sufi leaders change the way they talked around the middle of the 3rd/9th century?

Several features in this visualization relating to Melchert’s “transition” model that are worthy of note. The first is that a number of significant “leaders” that are not identified by al-Sulamī appear in this dataset; thus, the incorporation of earlier / other views about the history of Sufism substantially improves on Melchert’s methodological decision take al-Sulamī’s presentation of significant figures as the “convenient” yardstick by which to measure significant developments in the history of Sufism. Additionally, according to this data there appear to be no “stable” phases within the dynamics of early Sufi activity in which an “ascetic” or “mystical” lexicon predominate. Thus, this data offers no basis to suspect that a “phase transition” in the consistency of early Sufis’ use of ascetic and mystical terms occurred, in the way that Melchert has described. The data at the middle of the 3rd/9th century – that period which Melchert identifies as the locus of the proposed “transition” – are also of particular interest. In agreement with Melchert’s proposition, this period appears to be one of relative “inconsistency,” in which most figures use ascetic and mystical terms at roughly the same rate. By comparison, the 2nd/8th century appears to be a period of extremes, featuring both the “most ascetic” and “most mystical” figures in the entire dataset. The reason(s) for these features of the data is / are open to further conjecture, and we address one possibility in the forthcoming article.

Final Thoughts

There are no quick fixes in the effort to apply computational methods to notional entities, such as groups and (derivatively) their leadership. In order to present viable analyses, methods should also accord with Thiebault’s maxims for visualizations in being transparent, accurate, and information rich. Melchert’s attempt to use a primitive computational approach to model the dynamics of early Sufi behavior is commendable, but ultimately falls flat beause of his preference to put the model before the data. The weaknesses of Melchert’s method and empirical findings notwithstanding, the “transition” model he proposes looks suspect even when a dataset that better addresses these research design issues is used to test its fitness.

It’s my suspicion that this last issue – testing existing “scholarly commonplaces”– is where much of the value in computational approaches to history will lie in the near future. The effort to explain the dynamics of human behavior in the remote past has proceeded, especially in the field of Islamic Studies, according to the consideration of a single variable or dimension of the available data. New attempts in this endeavor will require the building of multi-dimensional datasets in order to demonstrate the applicability of weakly supported existing models, or in order to propose new models.11 Projects like Seshat are a step in the right direction on this front, although more granular data such as that presented in this post are desirable.

Endnotes

*: The bibliography is rapdily expanding, and I make no claims to comprehensiveness here. At least three separate journal have been dedicated to the relation of history and computation: History & Computing (Oxford: Oxford University Press for the Association for History and Computing, 1989-2006); and International Journal of Humanities & Arts Computing: A Journal of Digital Humanities (Edinburgh: Edinburgh University Press, 2007-), originally published under the title History & Computing (1998-2002); and Peter Turchin (ed.), Cliodynamics (2010 - ). doi: https://escholarship.org/uc/irows_cliodynamics. Numerous computational approaches are discussed in: Bojan Bozic, Gavin Mendel-Gleason, Christophe Debruyne, Declan O’Sullivan (eds.), Computational History and Data-Driven Humanities. Second IFIP WG 12.7 International Workshop, CHDDH 2016. Dublin, Ireland, May 25, 2016. Revised Selected Papers (Berlin: Springer, 2016) (link); and Lars Wieneke, Catherine Jones, Marten During, Florentina Armaselu, and René Leboutte (eds.), Proceedings of the Third Conference on Digital Humanities in Luxembourg with a Special Focus on Reading Historical Sources in the Digital Age. (Aachen: CEUR Workshop Proceedings, 2016).
A fairly detailed overview of applications from a social science perspective can be found in: John H Miller and Scott E. Page, Complex Adaptive Systems: An Introduction to Computational Models of Social Life, Princeton Series in Complexity (Princeton, NJ: PUP, 2008). A monograph heavier on equations is: Claudio Cioffi-Revilla, Introduction to Computational Social Science. Principles and Applications, 2nd ed. (Berlin: Springer, 2017) (link); and Juan A. Barceló and Florencia Del Castillo (eds.), Simulating Prehistoric and Ancient Worlds, Computational Social Sciences Series (Berlin: Springer, 2017). (link).
For a brief look at historical studies using computations derived through networks, see Tom Brughman’s bibliography; case studies are given in Tom Brughmans, Anna Collar and Fiona Coward (eds.), The Connected Past. Challenges to Network Studies in Archaeology and History (Oxford: OUP, 2016) (link), and Justin Leidwanger and Carl Knappet (eds.), Maritime Networks in the Ancient Mediterranean World (Cambridge: CUP, 2018).
There are also now several e-textbooks available on methods in computational approaches: Jack Dougherty and Kristen Nawrotzki (eds.), Writing History in the Digital Age (Ann Arbor, MI: University of Michigan Press, 2013); online version at doi: <www.digitalculture.org/books/writing-history-in-the-digital-age/>; and Lincoln A. Mullen, Computational Historical Thinking with Applications in R (2018-): doi: https://dh-r.lincolnmullen.com/ ↩︎

1: See, e.g.: Arianna Betti and Hein van den Berg, “Towards a Computational History of Ideas,” in Wieneke et al (eds.), Proceedings of the Third Conference on Digital Humanities in Luxembourg, doi: http://ceur-ws.org/Vol-1681/Betti_van_den_Berg_computational_history_of_ideas.pdf. ↩︎

2: Christopher Melchert, “The Transition from Asceticism to Mysticism at the Middle of the Ninth Century C.E.,” Studia Islamica 83 (1996): pp. 51-70. Further references to this article are provided in the body of the blog; all indications of emphasis are my own. ↩︎

3: For more about the titles attributed to these figures, refer to the Primary Sources page. ↩︎

4: For the problems associated with the postulation of “incommensurability” between phases after a “transition,” see: Yuri Mnyukh, “On the Phase Transitions that cannot Materialize,” American Journal of Condensed Matter Physics 4/1 (2014): pp. 1-12, at §10. (link) ↩︎

5: Bernd Radtke, “Von der hinderlichen Wirkungen der Extase und dem Wesen der Ignoranz,” pp. 251-292, at pp. 265ff. In Bernd Radtke, Neue kritische Gägne: Zu Stand und Aufgaben der Sufismforschung / New Critical Essays: On the Present State and Future Tasks of the Study of Sufism (Utrecht: M.Th. Houstma Stichting, 2005). Cf. Christopher Adair-Toteff, Fundamental Concepts in Max Weber’s Sociology of Religion (London: Palgrave MacMillan, 2015), pp. 55-82 (Ch. 4 - “Asceticism and Mysticism”). ↩︎

6: Sara Sviri, “Sufism: Reconsidering terms, definitions and processes in the formative period of Islamic mysticism,” pp. 17-34, at p.30. In Geneviève Gobillot and Jean-Jacques Thibon (eds.), Les Maîtres soufis et leurs disciples des IIIe/Ve siècles de l’hégire (IXe/XIe). Réflexions sur les concepts et l’évolution historique (Damascus and Paris: Presses de l’IFPO, 2012). ↩︎

7: Melchert may, in fact, have chosen the anecdotes used in the article in such a way as to assure their representativeness, although he does not state this. However, such a small n in his sample practically begs the question of whether a more robust examination of data would yield a result that differs from his. ↩︎

8: There is some uncertainty in this number, as it is not possible to positivly identify every named entity in the isnād s with a particular individual. For example, we have erred on the side of caution in considering an individual identified only as “Muḥammad b. al-Ḥasan” (a very common Islamic name), as separate from another figure named “Abū ʿAbd Allāh Muḥammad b. al-Ḥasan al-Baghdādī.” ↩︎

9: Paul Nwyai, Exégèse Coranique et Langage Mystique. Nouvel essai sur le lexique technique des mystiques musulmanes, Série I: Pensée Arabe et Musulmane Tome XLIX (Beirut: Dar El-Machreq, 1970). (link) ↩︎

10: See the documentation at https://camel.abudhabi.nyu.edu/madamira/. ↩︎

11: See, e.g.: Xavier Rubio-Campillo, “Model Selection in Historical Research Using Approximate Bayesian Computation,” PLOS One 11/1 (2016): e0146491. doi: 10.1371/journal.pone.0146491. (link)