* The pitch of the original objects could be determined in comparatively few cases; either bells that had an intact clapper and so involuntarily produced a note during normal museum handling, and objects whose pitch was documented by previous researchers, at a period when museum objects had fewer restrictions concerning handling. For the replicas, pitches are given as frequencies in Hz, and as approximations to the nearest tempered semitone using A440 as the standard. Bells produce more than one frequency, but usually with an identifiable dominant (primary) note, and sometimes lesser notes are also discernible (secondary, etc.). Crotals and cymbals clashed together produce multiple frequencies simultaneously, so a single value cannot be identified for these. Recording of replica frequencies took place outdoors on 25/06/2018, in weather conditions of 22°C and 55% humidity, using apps SpectrumView 2.2 and PitchAnalyzer 6.0. SpectrumView has the benefit of displaying the frequencies graphically and in relation to sound levels, so that it is possible to confirm which frequencies predominate in the case of instruments that produce multiple prominent frequencies.

4. Craft Construction: Ceramic Objects

The ceramic rattles in the Petrie collection are listed in Table 2 and are shown with their replicas in Figure 3a/3b, Figure 4a/4b and Figure 5a/5b. (It should be noted that rattles produce multiple frequencies simultaneously, so a single value cannot be identified, and this information is not included in the table.) The replicas were made by George Morris and Georgia Wright. The original objects have not been subject to compositional analysis. One of the objects, however, has its composition identified in the Petrie online catalogue as Nile silt, an alluvial clay that was commonly used for ceramic artefacts in Pre-Dynastic, Dynastic, Hellenistic, and Roman Egypt (UC65087). The clay composition of several artefacts manufactured from Nile silt has been investigated in scholarly research, and a commercial earthenware clay body was chosen that approximated to this composition (see Table 3) and used for all the object replicas. Grains of sand were visible in the clay matrix used for the original objects, and so a clay tempered with sand was chosen (10% 40/90 sand giving a medium-coarse texture). The pointed rattle (UC65087) also showed evidence of vegetable temper in the form of impressions left on the surface of the clay where an organic material had burned away during firing. Chopped straw temper was therefore also added to the clay used for the replica of this object in particular.

*HK1 is straw-tempered, HK2 untempered, see Redmount and Morgenstein 1996, table 1.

Test firings were used to establish the shrinkage rate of the chosen clay at the desired temperature (9.4%) and the 3D models were enlarged in size accordingly before printing in ABS thermoplastic. The chosen firing temperature was 700°. This is rather lower than the firing temperatures suggested for Nile silt clays (a minimum temperature of 750°: The Levantine Ceramics Project, see also Hamdan et al. 2014, 987). However, it was chosen based on the opinion of a craft practitioner who had examined the appearance of the original objects and who suggested that UC65087 in particular appeared to be rather under-fired (Eric Hall, pers. comm.). The qualities of test samples fired at a range of different temperatures were also taken into account.

Moulds were made for the objects in plaster of Paris using the ABS 3D prints coated in slip. For the nut-shaped rattle, with only one-half extant, two prints were made and joined using clay before the mould-making stage. Two-piece moulds were used for the replicas of UC71557 and UC65087, while for UC34972 a separate head part was also required in addition to the two main horizontal halves of the mould. A break in the original object, where the head is missing, shows that there was no vertical join and so this was felt to be the most likely original method, rather than the use of two moulds dividing the object vertically. It was evident from inspection of the originals that UC34972 and UC71557 were originally made in moulds. Although the original of UC65087 had probably been made by hand without the use of a mould, it was decided that using a mould would better preserve the dimensions of the original object, and so these were used for all of the ceramic objects.

As noted above, it was difficult to evaluate the wall thickness for the bird rattle and pointed rattle (UC34972 and UC65087) so the thickness of the clay used was based on the information from the nut-shaped rattle where the original was c. 3mm thick. The bird rattle appeared to be of a similar thickness, at least in the area around the hole in the base of the object. Slightly thicker clay (c. 5mm) was rolled out, which then was compressed slightly as it was pressed into the moulds, resulting in a thickness of approximately 3–4mm.

There was information from the bird rattle original concerning the appearance of the pellets inside, which were partly visible through the hole in the bottom of the object and, fortuitously (although not from the museum's perspective), one of these fell out when the object was being examined, allowing its shape to be noted before it was placed back inside. Similar small oval pellets about 0.5–0.8mm long were added to the halves of each mould before these were pressed together to make each complete object. Since there was no information available concerning the likely quantity of pellets inside each rattle, a number of different replicas were produced for each original using different quantities of pellets. Those used in sound recording are shown in Figures 3b, 4b and 5b.

The objects were finished by hand with the use of wooden tools to replicate the decorative detail used. The bird replicas were also coated in white slip as this had been used to coat the reddish clay body of the original object. The objects were then fired in a non-vented electric kiln on a drying - slow biscuit programme of 10° per hour for 15 hours, then 30° per hour to 700°.

5. Craft Construction: Wooden Objects

The two sets of wood clappers were the most straightforward to produce since each half could be cut out in wood by sending 3D scan data directly to a wood router, thus preserving the original dimensions of the objects very closely. Details are given in Table 4 (it should be noted that clappers produce multiple frequencies simultaneously, so a single value cannot be identified, and this information is not included in the table) and originals and replicas are shown in Figures 22a/22b and Figures 23a/23b. The replicas were made by Keith Greenhow, Daniel Knox and Julien Soosaipillai. The commonest wood used in the Roman period for small turned objects is European boxwood; however, this species is very slow growing, and so large pieces of timber rarely become available. We were not able to source large enough pieces to match the dimensions of the objects, and so a New World species with similar properties, castello boxwood, was used instead. The other question for this object type concerned whether the two halves of the clapper had originally been joined together and, if so, how. Similar modern objects used in Turkey are not joined (Kevin Dawe, pers. comm.) and this seemed most likely for the clapper set with a single large hole through each handle end (UC59603 and UC59604). Experimenting with the replicas of these objects suggested that tying the two parts together using the hole did not hold the objects in the correct tension for them to be used effectively held in one hand. It was probably used instead for a long, loose tie so that the objects did not become separated from one another. Conversely, the original of UC71305a–b possessed matching pairs of small holes drilled diagonally through the object which proved suitable for joining the two halves. Although the wood router was not capable of creating these holes in the original cut, their position was easy to identify from indentations on the replica objects and they were drilled through using a narrow drill bit similar in dimensions to that of the original holes. Drawing on examples of historical musical instruments such as castanets (e.g. those displayed in the Oxford University Bate Collection), catgut (usually made from the intestines of ruminant animals) was selected as a joining material. When it was threaded through the pairs of holes, joining the two parts together, a very effective set of clappers was produced that could be operated with one hand.

* Frequencies produced by each tube are given in Table 5.

6. Craft Construction and 3D Models, Reed Objects

The objects were printed in PLA, a biodegradable thermoplastic made from cornstarch. One set of panpipes was also SLA printed in photopolymer resin, which gives a more accurate result with, for instance, fewer external traces on the object of the printing process. However, there was no difference in the sounds produced by the PLA compared to the photopolymer set. The original of the double pipes was missing its separate mouthpiece parts and so the 3D print of the main sections of the flute could not be played directly. Replicas were also made in organic materials of both the double pipes and the panpipes, by Georgia Wright. Bamboo was used as a material as it proved difficult to source reeds that were sufficiently robust and of the correct diameter. Bamboo canes as close as possible to the correct external diameters were selected and the interiors were drilled out to match the diameters of the original instruments by selecting drill bits of appropriate size. The pipes were cut to the same lengths as the originals.

6.1 3D printed panpipes

Initially the 3D printed panpipes only made a sound from the highest tube, which on the original was blocked with beeswax, and so also had a sealed end in the 3D print (Figure 8a). Once all the tubes had been blocked with beeswax, it was possible to play a complete series of notes. Fortunately, there is corroborating evidence from textual sources that wax was used to tune pipes (Wardle 1981, 140; pseudo-Aristotelian Problems 19.23, trans. Forster 1927, 919) and so we can be certain that all the tubes would originally have been blocked at the ends. Closing the end of the tube doubles the length of the sound wave and so lowers the note played by an octave (Campbell and Greated 1987, 197). Since the notes played by the pipes depend on the length and diameter of the tubes, the thickness of wax used for each tube will affect the frequencies produced. We experimented with thicknesses between 0.5 and 5mm. Below 0.5mm the wax layer became very fragile and prone to damage, and above 5mm it became increasingly difficult to force wax up the tubes, especially those of the smallest diameters, even when softened by warming in the hands. (Our observations relate only to the replica of UC33270 created in PLA and may not be valid for larger instruments.) An alteration in length above 5mm could also be made more efficiently by trimming the length of the tube as a whole. We measured the frequencies of the pipes with different thicknesses of wax and from this it was possible to evaluate possible ranges of notes that each tube could play, and thus what musical scales appeared plausible. The series D E♭ F G A♭ B♭ C (where the lowest note is D6), two conjunct diatonic tetrachords, could be produced very easily by applying the same amount of wax to each tube, a thin layer up to c. 3mm. Other attunements are also possible, however (ongoing research by David Creese). Subsequent to making the replicas, new work on the panpipes probably from Tebtynis, including a CAT scan, showed that wax depths could sometimes exceed 5mm, albeit on larger pipes; see the conference paper by Avanzini et al. (2016).

The frequencies are shown in Table 5 and Figure 24. It is relatively easy to vary the note played by over-blowing or altering the mouth position, so when measuring the frequencies, the aim was to keep the perceived air pressure and mouth position consistent.

6.2 Bamboo double pipes

Finger-holes were drilled in the double pipes and the two separate tubes tied together with jute string. Since the original mouthpieces were missing, reference to comparative data was made in order to reconstruct these elements, namely, a more complete set of double pipes in the Petrie collection (UC35805, dated to the Islamic or Byzantine period, from Gurob) and more recent historical examples of a closely similar middle-eastern instrument, called a mijwiz or zummara, in the Oxford University Bate Collection (acc. no. X4063 from Beirut, Lebanon; acc. no. (VIII 122) X4070 from Galil, Israel). Evidence relating to the ancient Greek aulos in the Graeco-Roman period in Egypt suggests that this had a different type of reed mouthpiece (Hagel 2010, 71–73 and fig. 13), again confirming that our instrument and UC35805 are likely to be Islamic period. Online information on the creation of mouthpieces was also utilised; see for example 'How to Make the Traditional Reed Pipe "Zummara"' (Hirmer 2016). Using the comparative material as a guide, a mouthpiece was made from a short section of reed cut so that it was sealed at one end by the natural diaphragm inside the reed. A small notch was made near the open end of the reed section and a cut was made upwards from this, parallel to the axis of the reed, creating a vibrating tongue attached only at one end. The open end was then trimmed with a knife to fit into the diameter of the main section of the tube and the process repeated for the second tube. To play the pipes, virtually the whole of the mouthpieces must be enclosed inside the mouth. The original and replica are shown in Figure 9a–b.

6.3 Bamboo panpipes

The panpipes were cut to the correct dimensions as described above, and one side of each individual tube was bevelled using a sharp knife, as on the original set of pipes. They were fixed together with two split pieces of bamboo, and thread, similar to the originals. On the original, pitch appeared to have been used to reinforce the joins between the separate parts, and to protect the surface, which made it difficult to identify how thread had been wrapped around the pipes, although in places crossed thread could be seen. The technician experimented with different binding patterns and chose the one that was most effective in fixing the tubes in place, with crossed thread across each tube. Two sets were made, one fixed with jute twine and the other with waxed thread (Figure 8b shows the latter, the bamboo replica used in the sound recordings).

The PLA set was used for the main experimental work evaluating wax depths since it was judged easier to completely remove the wax from the PLA surface without damage, and PLA sets were also used for sound recording (see below regarding the timbre of the sound). The bamboo set, however, was indispensable for evaluating the possible differences to the notes made by the horizontal alignment of the pipes. Inspection of the original artefact, and the 3D print, which could be handled to a much greater extent, showed that the lowest tube was twisted by about 30° out of its original alignment. Since in the bamboo replica the tubes could be rotated, it was possible to check that the same note was produced when the tube was perfectly horizontal and when it was rotated to the same degree and in the same direction as the original. This indeed proved to be the case.

7. Discussion

The main parameter affecting sound levels and notes produced (frequencies) is the dimensions of the objects, and since the 3D scanning process is accurate in this regard, those instruments generated directly from 3D scan data without intervening processes will have the highest levels of accuracy regarding the sound power levels (decibel levels) and notes (frequencies) produced by the objects. This applies to the clappers made by use of the wood router, and also to the panpipes printed in PLA, although in the latter case we have to take into account that some measurements were reconstructed from caliper measurements of the original rather than obtained directly from scan data. We should also take into consideration that PLA is heat-sensitive, and so a very hot ambient temperature might have a distorting effect on the notes produced by the replica artefact compared to the original, although since it is made from one material we can envisage that it would expand consistently, and so the intervals would be preserved even if the pitches were slightly affected. Temperature increases would also potentially affect the wax plugs. A consistent tuning, however, was maintained by the instrument across an ambient temperature range of 17°C to 28°C during a very hot UK summer in 2018. (The temperature of the air inside the tubes will also affect the tuning, for instance across this temperature range, sharpening the pitch by about a third of a semitone, but this effect will be uniform across all the tubes.) An SLA print in photopolymer resin, which is not affected by heat, could be used if changes to the dimensions of the PLA, because of the temperature, were of concern in the environment in which the replica was to be used.

The next most accurate in dimensions will be those bells and cymbals that did not need any alteration in the 3D model in order to make a successful casting from a 3D wax print (see Table 1 for details of measurements). These can also be suggested to have decibel and frequency levels that are likely to be accurate.

The three smallest bell replicas (plus that of the rumbler bell), and one cymbal, where the 3D model was thickened by up to 0.5mm to achieve a successful casting, vary from the originals in frequency. We can be certain about this, since the originals of the bells on bracelets were in such good condition that they produced sounds when moved during inspection, which were recorded, and the frequencies later assessed from the recordings. In one case, UC58540, the notes produced by the original and the replica were only a tone apart in the same octave, but in the other cases there was more divergence (see Table 1), probably exacerbated by the very small size of the bells, which means that a slight alteration of the dimensions has a proportionately greater effect than it would in a larger object. The adjustments to the models prior to casting will also have had an effect on decibel levels, although this is difficult to evaluate since we have no information on the decibel levels of the originals. We can perhaps make a very approximate estimate by comparing the decibel levels produced by two replica bells on bracelets with dimensions close to each other, both with a 17mm inner diameter, but slightly differing in height and wall thickness (UC58536, height of bell not including loop 18mm; maximum wall thickness at edge 1.2mm; UC58540, height of bell not including loop 21mm; maximum wall thickness at edge 1.4mm). They produce decibel levels within a 5 DB(A) range (see Swift et al. forthcoming, appendix on sound recording). A similar variance is thus very plausible for our originals of UC58536, UC58538, and UC58540, with walls thickened by 0.5mm, compared to our replica objects. We can emphasise that all would still be perceived as very quiet objects in general.

The forged cymbals diverged in diameter size from the originals by up to 1mm. This small difference in dimensions between the originals and replicas will affect the decibel levels to some degree. There are noticeable differences in loudness between the smallest replica cymbals (UC35797) and the largest ones, with a diameter 7mm larger, but a discrepancy of 1mm will have much less effect. The forging process may also have introduced variance in pitch between originals and replicas (original pitches could not be recorded because of museum handling rules). The finished replicas of UC35797, for example, produced notes a semitone apart when struck at one individual point with a stick (see Table 1), although the same size blanks were used. When clashed together, however (their most likely mode of use to judge from visual sources, see Figure 18), multiple points on the edges of the cymbals make contact with each other, producing multiple frequencies (see Figure 20). Thus we can envisage they were not intended to produce a single dominant frequency. Small pitch variations evident when each object is individually struck are therefore not significant.

The ceramic objects will be the least accurate in both frequency range and decibel levels, since the firing process affects the size of the objects, and we have little information about wall thickness, and none on the number of pellets that would have been included inside each rattle. To address the latter problem, a number of replicas were made for each original, with different numbers of pellets inside. For the bird rattle, that chosen for use in sound recording was the replica closest in weight to the original (see Table 2). For the pointed rattle, the original was intact and made a noise when moved during inspection, so this sound was recorded, and then the closest match in sound was selected from the replica rattles, even though this rattle had a greater discrepancy in weight with the original compared to some of the other versions. Despite the possible variance introduced by shrinkage during firing (compensated for by enlarging the original 3D models), the size ranges of the objects are very close to the originals. Wall thickness for all three rattles was based on that of the nut-shaped rattle, where the original was broken and thus could be measured, so this rattle replica will be most accurate in terms of decibel levels. The wall thickness of the original bird rattle looked similar to this from inspection of the hole underneath, although wall thickness might vary elsewhere on the object. Since each rattle is clearly distinguishable by size (see Table 2 and Figures 3a/3b, Figure 4a/4b and Figure 5a/5b), we can envisage that the relative loudness of the instruments as compared to each other will be the same, although the accuracy in the decibel levels may vary. Rattles simultaneously produce a range of frequencies that the human ear cannot distinguish from each other, so differences in frequency range will not be especially significant. It is also worth making the point that ancient objects like this, probably not intended for music performance (unlike our other instruments such as the panpipes) will have had a good deal of variance from one another, and even if we have not matched the decibel levels and frequency ranges of our original rattles exactly, it is likely that some Roman rattles once existed with the same qualities as the ones we have produced.

The different materials that our replicas are made from compared to the originals will not affect the decibel levels and pitches (frequencies) produced, but they will affect tone quality. Tone quality can only be subjectively assessed in any case, and so has not formed a part of the strictly experimental investigation, but it does of course affect our perceptions of the sounds in the recordings of the objects (discussed further below) and so it is worth considering briefly. How much difference in tone quality would be evident between clappers made of different types of wood, ceramic objects made in similar earthenware fabrics but with slightly different recipes, or different brass alloys, is debateable. Differences in materials are more marked for the cast metal bells between original and replica, but here we have been able to choose an alloy with similar acoustic properties to three of the bells replicated. The object that is most different from the original in terms of materials is the panpipes. Although replicas were made in organic materials, since the 3D prints in plastic best preserved the exact dimensions of the object, one of these, a PLA print, was used for the decibel and frequency measurements and the sound recordings (the most accurate is the photopolymer print, but unfortunately this was not available at time of recording). How close is the sound quality of this 3D print to a reed original? Scholars of acoustics take the view that different materials cause minimal differences in sound quality for wind instruments, as the vibrating air column is the most important part for producing the sound (Campbell and Greated 1987, 295, 404–7). To evaluate this for PLA compared to organic plant material, we 3D scanned and printed a modern whistle made from a plant stem (of unknown type) and compared the sound produced by the whistle and the 3D print replica. We could not audibly distinguish any differences between them. Similarly, the replicas in bamboo made of the PLA pipes sounded virtually the same, perhaps slightly more breathy in tone although this may be a subjective judgement. A project that created a replica of a 16th-century recorder through 3D scanning and printing also found that a professional musician could produce a tone quality from a PLA replica the same as that of the wooden original (Witkowski 2017).

Once the process of replica creation had been completed, sound measurements of the decibel and frequency levels were made and used in an evaluation of the likely social function of the objects. Since this article focuses on the replica creation process, it is not the place for a detailed discussion of the social interpretation of the sounds made by the objects, but we can give a specific example to illustrate the usefulness of the information. Measurements of decibel levels, for instance, confirmed that some instruments such as the clappers, cymbals and panpipes, could have been heard by a wide audience when used outdoors and so would be suitable for public music performances (as indeed is suggested by textual and visual evidence) while others, such as the nut-shaped rattle and bells on bracelets, were very quiet and so are likely to have been for personal use in the domestic context only (see Swift et al. forthcoming for full details).

8. Sound Recording

8.1 Recording process

Sound recordings of the objects were made in a professional quality sound studio at the University of Kent by sound technician Frank Walker. The acoustics were modelled virtually, using the dimensions of a courtyard (approximately 2.5 × 11.8m) from a Romano-Egyptian house at Tebtynis (Figure 25), House 3200-III Room F (Hadji-Minaglou 2007, 117–26 and fig. 56). This was chosen as a typical domestic setting in which the instruments are likely to have been used (see Swift et al. forthcoming). The music performers were myself, the other project team members April Pudsey and Jo Stoner, and volunteers Ada Nifosi and David Walsh. Levels of musical knowledge varied from none at all, to ability to read music and some experience in playing and singing in amateur performances. We deliberately chose not to involve professionally trained musicians or music students, since we wanted to record the sound of instruments as played in everyday circumstances, which would have included amateur performers.

Recordings were made using two microphones set at different distances from the player of the instrument so that two recordings, one at close range and one at a mid-range distance, could be made simultaneously. Most of the instruments were deliberately agitated in order to produce their typical noise. The decibel range for each one was also recorded, by playing from as softly to as loudly as possible. The cymbals produced different noises according to how they were struck and/or held and so a range of recordings were made to capture the various possibilities. For those instruments where other uses were likely, these were also recorded, namely, the passive sound that the bells on bracelets attached to the wrist would make on movement, and the sound of the rattles when rolled on the floor as well as when shaken deliberately. Representative recordings are available in Appendix 2.

8.2 Tunes, rhythms, and ensembles

A selection of the artefacts (the wooden clappers and cymbals, the bird rattle, and the panpipes) were recorded playing specific tunes or phrases and rhythms, which were selected from scholarly publications with transcriptions of ancient music in modern musical notation (Pöhlmann 1970; Pöhlmann and West 2001; West 1992). Two sets of panpipes were tuned to different scales to reflect the variation available from this object: in addition to the diatonic scale described above used for the first set (D E♭ F G A♭ B♭ C), a further diatonic tuning was used for the second set (D E F# G A B C). Berlin Papyrus 6870 line 13 (Pöhlmann 1970, cat. no. 31; Pöhlmann and West 2001, cat. no. 51) was chosen as a tune playable by the first set, and one of the tunes from Bellermann's Anonymous (Pöhlmann 1970, cat. no. 12; Pöhlmann and West 2001, cat. no. 37) playable by the second set; both are from the Roman period. Rhythm patterns were selected from West 1992, table 5.1, choosing those within a chronological range that included the Roman period. For the cymbals and clappers, instruments which, from visual sources, we can suggest were often used by dancers (Hickmann 1949b, 523–30; Musée d'archéologie méditerranée Marseille 1997, 238), we used five-beat rhythms that were popular for dancing (West 1992, 140–43). A different rhythm also used in the Roman period was chosen for the bird rattle, an Ionic rhythm, with three beats in the bar (West 1992, 145–7), to introduce some variety (the loudness of this rattle suggests that, unlike the other rattles, it could be suitable for music performance). Recordings were also made of two percussion instruments (a set of clappers, and cymbals on handles) accompanying firstly the panpipes tune for the first set of pipes, and secondly a Roman-period song, the 'Song of Seikilos' for which the words and music are both extant (Pöhlmann 1970, cat. no. 18; Pöhlmann and West 2001, cat. no. 23). Visual sources confirm that cymbals on handles were sometimes used to accompany panpipes (Hickmann 1949b, 474; for a specific example, see the Carthage banquet mosaic, Dunbabin 2003, fig. 46). Finally, all the sound recording participants selected one or more instruments to play together freely when processing past the microphone, in order to mimic the sound of a procession passing by.

9. Public Engagement

Sound recordings and replicas were also used for public engagement purposes in a temporary exhibition at the Petrie Museum of Egyptian Archaeology, UCL. A selection of the less valuable replicas and 3D prints were made available throughout the exhibition (3D prints of panpipes and clappers, and craft replicas of ceramic rattles, panpipes and clappers), with handling of the other replicas, mostly the metal ones, available at four workshops for the general public. Activity sheets were available in the museum with guidance on how to play the instruments (for instance manipulating them in different ways to produce different sounds), and examples of tunes and rhythms that visitors could try out. These were also used in the public workshops.

3D models of some objects, that could be viewed and rotated by visitors, were also included in the exhibition, together with sound recordings (as described above; see also Appendix 2) to give visitors a further idea of the sounds produced by the objects, the kinds of ensembles in which they might be played, and examples of authentic tunes and rhythms. Each sound clip could be accessed by clicking on a picture of the relevant instrument on a computer screen. After the image had been selected, the sound, delivered through headphones, was preceded by a few lines of text on the screen setting out an imaginative scenario, to encourage the listener to consider the sound in the context of everyday life in the ancient world. For example, the sound clip of the panpipes playing the tune from Bellermann's Anonymous (see above), which is one of a set of Roman instrumental exercises for learners, was preceded by the text 'Sarapion, a young boy, is learning to play the panpipes. Listen as he practices in the courtyard'. Research findings from the experimental archaeology undertaken using the replicas were also incorporated into the text panels for the display cases in the exhibition, for instance information about how quiet some of the objects were.

A total of 115 visitors filled in a visitor questionnaire about their experience of the exhibition and its associated activities (see Appendix 3 for the blank form). The completed questionnaires covered a broad spectrum of ages, with similar numbers across the different age categories (Figure 26); 38% of the respondents were in higher education. No particular trends were evident in the data for the individual age categories, and so it was examined as a single dataset.

The sensory experience of the sound, and how this helped to bring the past to life, and make it more relatable for visitors, was the strongest theme of the feedback (see Table 6). Western culture is a predominantly visual one, and because museum visitors cannot usually handle original artefacts in museums, their experience of ancient objects is also primarily visual. Using objects that were both audible and tactile, and listening to sound recordings, facilitated a more immediate experience, in which people felt more connected to the past, and were consequently better able to imagine people using the original objects within a living social environment ('they seem more realistic as humans [rather] than just history').

We asked visitors to indicate what they liked best about the exhibition in a multiple choice question (Figure 27). Was it seeing the original objects, trying out the replicas, hearing the sounds, or playing with the 3D models of the instruments? Most found it difficult to choose, with more than half choosing multiple aspects. The most popular individual choice, of about a quarter of respondents, was hearing the sounds. Those who liked multiple aspects also mentioned the sounds more often than the other elements. Although trying out the replica instruments was selected as the individual favourite by comparatively few (6 respondents), the sound recordings also depended on the existence of the replicas, and so we can conclude that the artefact replicas were fundamental to the popularity and success of the exhibition. In the section 'Tell us more about what you liked', multiple respondents also mentioned hearing or experiencing the sounds (14), enjoyment and/or fun (8), and the opportunity to see the original artefacts (6). From the multiple choice question, a combination of seeing the original artefacts and hearing the sounds produced by the replicas was popular (13 respondents), showing that the replicas and 3D models were not seen as a substitute for viewing the originals, but rather, as a supplement that enhanced visitor experience of the original artefacts.

In response to the question 'What did you learn that was new?', 56% of respondents said that they learned something new about music, sounds, or instruments. They reported learning what instruments were played, and how to play them. They said they had learned what instruments sounded like and how loud or soft they were. These aspects were all directly communicated by the artefact replicas, and the sound recordings made using the replicas. Also mentioned were rhythms, musical notation, soundscapes, how the instruments were made, and how music contributed to everyday life.

Another common theme was how the exhibition had broadened people's conception of Egypt. Sixty-two per cent of visitors said that it had changed their view of Ancient Egypt. Some did not have much or any awareness of the Roman period ('I hadn't thought about Romans being in Egypt'; 'previously I had little sense of a Roman culture in Egypt'). For others, who had a popular image of Egypt as being mostly about things like burials and hieroglyphs, it broadened their conception to include aspects of everyday life ( 'Oftentimes we hear of Egypt in terms of elaborate tombs, and this showed the life of the everyday'; 'Nice to listen to Egyptian sounds rather than just funeral objects/graves'). Many people had not previously thought about music and sound in the ancient world at all (mentioned in 7 responses), demonstrating the value of the exhibition in transforming wider perceptions of the past.

10. Conclusion

To summarise, the processes of 3D scanning, 3D model creation, 3D printing, and craft reconstruction described above can be used to create a set of replica objects that can be described as 'functional replicas' – that is, they replicate with reasonable veracity, greater for some objects than others, the features of the objects under investigation. They can be used for both research and public engagement activities. From a research perspective, the accurate re-creation of decibel levels and pitches for some objects and, less precisely, their approximate timbre, has facilitated an enhanced understanding of the likely social context of use and audibility to users within a wider acoustic environment of a range of sound-making objects. Visitor feedback from the use of the artefact replicas in the 'Sounds of Roman Egypt' exhibition at the UCL Petrie Museum makes clear their strong contribution to visitor experience, particularly with regard to creating a sense of immediacy for visitors that helped to bring the past to life. We have also been able to evaluate to a certain extent how far the necessary compromises described above have affected the sounds produced by the objects since in a few cases we have both original and replica sounds to compare with each other. Small variations in measurements were found to significantly affect pitch, and so this will be exact in replica creation only where no variation has been introduced in the manufacturing process. 3D scan data and sound recordings have also been made available for future use in both research and public engagement. Finally the re-creation of a plausible acoustic environment, the courtyard of a Romano-Egyptian house of a type in which many of the instruments are likely to have been played, provides an insight into the acoustics of the space and its effect on the sounds of the instruments, and enables us to consider wider aspects of the experience of soundscapes in Roman Egypt.

Acknowledgements

The project for which this is one of the published outputs, Roman and Late Antique Artefacts from Egypt: Understanding Society and Culture, was funded by an AHRC project grant, and we are very grateful to the AHRC for this support. Thanks also to current and former staff at the Petrie Museum of Egyptian Archaeology, especially Louise Bascombe, Anna Garnett, Maria Ragan, Alice Stevenson, and Alice Williams, and to the Petrie Museum and the British Museum for permissions to reproduce images of objects in their collection. David Walsh and Ada Nifosi kindly assisted by playing some instruments for sound recording, and Egert Pöhlmann by allowing us to use his transcripts of music in the recordings. Eric Hall, formerly a professional potter, kindly provided advice about the replica creation for the ceramic artefacts, and Prof. Kevin Dawe, University of Kent, and Andrew Lamb of the Oxford Bate Collection, assisted with information about comparative ethnographic musical instruments; Keith Greenhow, Daniel Knox and Julien Soosaipillai (all University of Kent) made the replicas of the wooden clappers (other replicas were made by the technicians and craft practitioners credited above). Many thanks to all of them.