Introduction & Background
This paper details a computational analysis of sacred music, comparing Jewish and Muslim prayers from Morocco, seeking an answer to the question of whether geography or faith hold more influence over melodic expression. This study began as a comparison of Ashkenazi and Sephardi Jewish prayers to see if there were discernable patterns among their similarities and differences, and whether or not those connections could be based in some way on geographical origin (Ashkenazi traditions stem from Eastern Europe whereas Sephardi Jews come from North Africa). I qualitatively observed similarities between the Sephardi recitations and Arab musical traditions of North Africa which both employ similar musical flourishes and embellishments beyond vibrato. These stylistic similarities intrigued me, as did the concept of geography influencing the expression of sacred music, especially between two faiths with a long history of animosity toward each other. My goal in this study has been to demonstrate musical similarity and in so doing promote mutual understanding and empathy between Jews and Muslims of North Africa.
Source Material
To begin, I needed to choose prayers from each faith to analyze. The choice of Muslim prayers was simple, as the call to prayer, or the Adhan (also sometimes transliterated as azan), was what sounded reminiscent to me in the Sephardi samples I had previously analyzed. As for the Jewish prayer, initially, I began with the prayer I had used in the previous study, the Kiddush (קידוש) which is used to bless wine on holidays. At first, I thought that a frequently used prayer would generate more results and provide an abundance of examples to choose from in order to find quality samples, however, I encountered difficulty finding versions that were textually identical. After learning from a Rabbi that the Kiddush is recited differently for each holiday, he suggested that I instead search for biblical verses because they have no textual variation. He recommended I use a prayer known as the Shema, or Shema Y’Israel (שמע ישראל) [Hirschfield, 2021]. Interestingly, in the same conversation, the Rabbi also told me that the word Adhan comes from the Arabic root, “to listen,” which he said triggered an association with the Shema because it has some parallels to the Shehada which opens the Adhan. In future studies, I would like to investigate more closely these textual similarities, however, the focus of this particular study was musical analysis.
Samples were sourced from YouTube using the search terms שְׁמַע יִשְׂרָאֵל מָרוֹקוֹ (‘Shema Y’Israel Morocco’) and أَذَانأَذَان المغرب (azan of Morocco). In the case of the Shema, I started my search using English terminology and was unable to find many good examples. Switching to Hebrew provided much better results and I was able to find four examples. Learning from this experience, I similarly searched for Adhan samples using Arabic search terms. This provided ample search results, however, the majority of examples were so heavily laden with reverb and delay that it would have made analysis extremely challenging if not impossible. I imagine this is because Adhan is classically recited from the tops of mosques, intended to be heard from a great distance, and is, therefore, part of the style [Albakri 2019]. While there are a number of plugins that can remove reverb, most are not very effective, and the few that work well are prohibitively expensive. So, I painstakingly listened through dozens of videos until I found four that had a manageable amount of reverb.
I ended up with eight samples for analysis; four examples of the Shema, and four of the Adhan, all of which came from Moroccan traditions. It was important to me to study a few versions of each prayer to see if there were any discernable patterns among the samples of the same prayers before comparing them across faiths.
Analysis
There was a great deal of preprocessing necessary to prepare samples for analysis and to ensure accurate results by minimizing chances for outlying data points caused by recording artifacts or poor recording quality. This processing was done in Ableton Live using FabFilter’s Pro-Q3 plugin to attenuate frequencies below around 400 Hz and above 5k Hz using “brickwall” high and lowcut filters respectively. Without this step, the low and high frequencies that were not related to the primary melody would have confused the pitch tracking algorithms. I additionally needed to put individual instances of the same EQ plugin on each of the tracks as they had unique problems that needed to be filtered out (see figure 3). For example, one of the clips had the sound of crickets in the background, which would have detracted from the primary melody.
After EQs had been applied to all of the samples, I then arranged and parsed the audio according to lines of text such that each phrase was separated and aligned (see figure 1) to ensure I was comparing the same line of text among each of the samples. I repeated this process for all of the samples of the Adhan as well as the Shema. After the clips had been consolidated and grouped according to lines of text, I selected a line that had the most intriguing melodic variation shared by all of the samples. In the case of the Shema, I chose the line beginning V’ha-yu ha-d’va-rim ha-ei-leh, or, וְהָיוּ הַדְּבָרִים הָאֵלֶּה in Hebrew, which means, “And let these words…” [Hirschfield, 2021]. The pitch variations among the examples of the Shema are particularly interesting because the sacred text of Judaism, the Torah, includes notation for how the words should be sung. These notations are referred to as ‘Torah trope,’ or cantillation, which do not indicate specific pitches, but rather, compliment vowels to indicate how text should be chanted [Van Kranenburg et all, 2011]. For the Adhan, I chose the line Hayya ‘alas-salat, or, حي على الصلاة in Arabic [Alaqad, 2008]. Every line in the Adhan is repeated twice, and I observed in the samples that the second recitation had more elaborate melodies. In research, I discovered that while there is no formal rule that details this explicitly, it is a widely accepted interpretation that the second verse is an invitation for musical elaboration [Frishkopf, 2018]. Furthermore, there is no textual direction for how the Adhan should be performed, so melodies, timbre, and rhythm are entirely left to the mu’adhin, or the person who performs the call to prayer [Frishkopf, 2018]. After completing the preprocessing, I then exported all of the samples as individual .wav files and imported them one by one into another application, Sonic Visualiser, for analysis.
In Sonic Visualiser, I used Vamp plugins and applied a pitch transform called, “pYin: Smoothed Pitch Track” (see figure 4), and exported the annotation layer to .csv using the following parameters: Yin threshold distribution = Beta (mean 0.15), suppress low amplitude pitch estimates (0.1000), onset sensitivity (0.7000), duration pruning threshold (0.1000) (see figure 5). These are the default parameters suggested by Sonic Visualiser because they work well with most monophonic audio. In future studies, I could adjust these settings to look more closely or broadly at shifts between notes, for example, to exclude the pitch shifts associated with vibrato as a variable.
Next, I imported the .csv files into a spreadsheet to analyze the data. For each sample, I ended up with one column marking time, and another column for value which related to frequencies present at that moment in time. I then created line charts to visualize the relationships between the samples in order to compare their pitch contours. While my preprocessing served me well and the lines were easy to follow which indicated isolation of target frequencies, for the initial analysis, samples were not time-stretched (in Ableton, this is referred to as warping, and warp mode was deselected on all of the clips in the first pass) because I did not want to impact the tonality of the samples by doing so. By not accounting for differences in timing, however, some samples were recited more quickly than others, so the pitch contours did not correspond to the same text. In order for the analysis of pitch contours to be accurate, the text needed to align perfectly, word for word so that I would be viewing the differences in pitch among the different versions of the same text. So, to achieve greater accuracy in graphical representations of pitch contours, I then time-stretched the samples using the “Complex Pro” warp mode in Ableton which preserves the clip’s original formants, allowing for increased manipulation without losing tonal quality [Ableton, 2021]. I also needed to normalize the starting pitches by taking the lowest value among the starting pitches and subtracting that value from each of the other starting pitches and applying the difference to the rest of the data sets. This, in a way, served a similar function as transposition, allowing all of the samples to be considered according to their pitch variations and contours rather than the specific pitches themselves which were not relevant to the study.
Results
Among the Shema samples, time-stretching to align the text revealed a marked similarity in melody with variations in melodic ornamentation, particularly at the end of phrases (see figure 6). However, even after time-stretching the Adhan examples to the same length, the phrasing was still very different among them, with some of the samples saying the first two words quickly and then holding notes toward the end, while others held the first word and notes longer, finishing quickly. This, as it turns out, is also a stylistic choice left to the mu’adhin [Frishkopf, 2018]. This meant that the pitch contours were difficult to compare to each other because the text did not align. I returned to Ableton and began to manually drag formants so that the phrasing would match, however, the timing differences were so significant between some of the samples that even when using the “Complex Pro” warp mode, the audio was stretched so far that there was dramatic degradation in the tonality, making the pitch unreliable to track due to audio artifacts from the extreme stretching. In future studies, I would like to further break down phrases into words, aligning each word to account for timing differences, and then see how this changes the results.
In addition to line charts, I also created histograms for each of the samples and then compared them to each other (see figure 7). The advantage to using histograms instead of line charts is that it allowed me to see the distribution of frequencies, one metric for describing the musical ornamentations that are audibly present in both the Shema samples as well as the Adhan samples.
Conclusion
I was able to demonstrate similarities and differences among the versions of the Shema, which can be seen clearly in the line chart below (see figure 6). All of the samples followed the same general melodic arc, diverging toward the end of the phrase, with two samples going up in pitch, and the other two going down in pitch, and all of the samples returning to roughly the same starting pitch. While the similarities between the recitations of the two prayers are apparent through listening, further analysis is needed before drawing definitive statistical conclusions. My primary issues were removing timing and phrasing as variables, as well as finding the right metric to demonstrate stylistic similarity. While one can listen to the samples and immediately hear their resemblance, demonstrating this with data is not as straightforward. It is, however, a fascinating comparison and one that I have not found evidence of being researched before. Toward the end of my work on this project, I discovered a line of inquiry related to Music Information Retrieval (MIR) which I intend to explore at a later stage [Jesser, 1991]. In conclusion, while this study revealed many intriguing avenues to pursue in future work, it did not answer the question of whether geography or faith hold greater influence over musical expression because the Jewish and Muslim samples were not compared with each other, only to multiple examples of themselves, because I was unable to find a singular metric that would illustrate the musical similarity between two groups with textually dissimilar origins. After discovering the Jesser 1991 study which outlines thirty-seven measures for musical similarity, and reflecting on the analysis of pitch contour that I completed, the next step in this project is to implement other statistical MIR measures on the same samples in order to make the comparison feasible.
Figure 1
The top image shows the arrangement view of an Ableton project with textually aligned samples of the Adhan, and the bottom image shows the same but for the Shema
Figure 3
This image demonstrates the EQ settings used on one of the Adhan clips in Ableton Live
Figure 4
One of the samples after the P-Yin Smoothed Pitch Track transform had been applied
Figure 5
Export settings used in Sonic Visualiser
Figure 6
The above chart shows time vs pitch of the time-stretched Shema samples, and the below chart shows the same but for the Adhan samples.
Figure 7
Above: combined histogram of the Shema samples, Below: combined histogram of the Adhan samples
Works Cited
Ableton – 9. Audio Clips, Tempo, and Warping – 9.3.6 Complex Pro Mode <https://www.ableton.com/en/manual/audio-clips-tempo-and-warping/>
Alaqad. “Text of the Adhan.” Islamic Forum, in Learn Arabic, published 23 May 2008, accessed 15 Nov 2021 <https://www.gawaher.com/topic/292599-text-of-the-Adhan/>
Albakri, Tsonka and Mohammed Mallah. “Al Adhan: documenting historical background, practice rules, and musicological features of the Muslim call for prayer in Hashemite Kingdom of Jordan.” Musicologica Brunensia 2019, 54 (1): 167-185, DOI:10.5817/MB2019-1-12
Frishkopf, Michael. “Paralinguistic Ramification of Language Performance in Islamic Ritual.” Yale Journal of Music and Religion 2018, 4 (1): 5-27, DOI: 10.17132/2377-231X.1099
Hirschfield, Bradley. Email to the author, Adi Moskowitz. 15 Nov 2021
Jesser, B. 1991. Interaktive Melodieanalyse, Vol. 12, Bern: Peter Lang
Van Kranenburg, Peter et al. “A Computational Investigation of Melodic Contour Stability in Jewish Torah Trope Performance Traditions.” Proceedings of the 12th International Society for Music Information Retrieval Conference, ISMIR 2011, Miami, Florida, USA, October 24-28, 2011
Data charts:
<https://docs.google.com/spreadsheets/d/1i-nXwdX3IJnzvj_735JSBu1W34k-JIEbCXyX88U5Q7U/edit?usp=sharing>
Audio Sources:
“أجمل الأذان الفجر في المغرب” (translation: The most beautiful morning call to prayer in Morocco) Uploaded by IbrahimAbuAshraf, 16 Aug 2013. Accessed 24 Nov 2021. <https://youtu.be/rKQqnpovOcg>
“اذان تخشع له القلوب بصوت جميل للمؤذن محمد قصاص” (translation: Azan the hearts are humbled by the beautiful voice of the muezzin Muhammad Qassas) Uploaded by الفرقان – Al Furqan, 12 Dec 2020. Accessed 24 Nov 2021. <https://youtu.be/qEQW9GUzAIs>
“أذان إبداعي لصلاة العشاء للمؤذن محمد بن مروان قصاص” (translation: Creative call to prayer for the evening prayer for the muezzin Muhammad bin Marwan Qassas) Uploaded by قناة القران الكريم Mar 28, 2020. Accessed 24 Nov 2021. <https://youtu.be/h4CHgpABNj4>
“آذان المغرب بصوت الشيخ أمين كعسيس” (translation: Azan of Morocco with the voice of Sheikh Amin Kasis) Uploaded by El Djazairia One, 7 May 2019. Accessed 15 Nov 2021. <https://youtu.be/w-vPQ41LsYU>
“أذان المغرب الشيخ عبد الباسط عبد الصمد” (translation: The call to prayer of Morocco, Sheikh Abdul Basit Abdul Samad) Uploaded by أحمد جمبو, August 12, 2017. Accessed 15 Nov 2021. <https://youtu.be/1ZcJbtoXfSI>
“הדגמת אמירת קריאת שמע בצורה נכונה נוסח מרוקאי“ (translation: ‘Demonstrating the correct reading of an audio reading Moroccan version’) Uploaded by דלק לנשמה – חנניה מזרחי, 29 Apr 2020. Accessed 24 Nov 2021. <https://youtu.be/F23QkYEEwgQ>
“Tefila תפלה – Shema Yisrael שמע ישראל (Moroccan מרוקאי) – R. David Kadoch” Uploaded by Darké Abotenou, 19 Dec 2013. Accessed 16 Nov 2021. <https://youtu.be/awlyL0o7ej4>
“Keriat Shema – Shema Yisrael – קריאת שמע נוסח מרוקו” Uploaded by David Kadoch, 16 Nov 2017. Accessed 16 Nov 2021. <https://youtu.be/lhWWK-a3ZJY>
“קריאת שמע בנוסח מרוקאי“ (translation: ‘Moroccan-style audio reading’) Uploaded by איתמר מלכא, Dec 21, 2011. Accessed 16 Nov 2021. <https://youtu.be/ZGjuboaN5DA>