Tag Archives: speech

Stonehenge Lego

What did the different parts of Stonehenge do to the sound?

Our 1:12 physical scale model of Stonehenge allows us to examine what different parts of the structure do to the sound. We can add and remove bits and see what difference they make. One of the intriguing things about Stonehenge is it had different configurations. This blog focusses on the configurations from the middle of the 3rd millenium BC. (We haven’t examined the acoustics of the original giant stone circle from circa 3000 BC, because even at 1:12 scale it’s too big to fit into our test chambers!)

Fig. 1. One of the configurations tested. All the bluestones have been removed (you can see where they were from the black crosses on the boards) and also the lintels from the outer sarsens. The source is in the middle, and the microphone to the right.

The most up-to-date archaeological understanding of the different Stonehenge configurations comes from Professor Tim Darvill and collaborators [1]. The list below summarises their proposed sequence of events, focussing on the ones that are most significant for the acoustics within the circle. Note, the exact sequence of some of these events is not known. The full model we mostly tested and reported in early blogs is the last one from circa 2,200 BC.

2620-2480 BC

  • Trilithon horseshoe comprising five sarsen trilithons set up in the centre.
  • Double bluestone circle of between 50 and 80 bluestones set up outside the trilithon horseshoe.
  • Sarsen circle comprising 30 shaped uprights linked by 30 lintels built outside the double bluestone circle.

2480–2280 BC

  • Central bluestone circle created within the trilithon horseshoe.

2280–2020 BC

  • Central bluestone circle and double bluestone circle dismantled and re-built as bluestone oval and outer bluestone circle.

What we measured

We took our scale model apart in stages to explore the effects of different components as shown in the list below. This did not following the exact sequence shown in Darvill’s paper because we didn’t have enough time. Nevertheless, this allows us to examine what the different components are doing to the sound. In each configuration we had the source in the centre and measured the same six microphone positions (see Figure 2). I have written previously about the measurement method.

Configuration A: Full model circa 2,200 BC

  • Sarsen trilithon horseshoe
  • Outer sarsen uprights
  • Outer sarsen lintels
  • Outer bluestone circle
  • Inner bluestone oval

Configuration B: No inner bluestone oval

  • Sarsen trilithon horseshoe
  • Outer sarsen uprights
  • Outer sarsen lintels
  • Outer bluestone circle

Configuration C: No bluestones

  • Sarsen trilithon horseshoe
  • Outer sarsen uprights
  • Outer sarsen lintels

Configuration D: No outer sarsen lintels

  • Sarsen trilithon horseshoe
  • Outer sarsen uprights

Configuration E: 5 sarsen trilithon horseshoe

  • Sarsen trilithon horseshoe
Fig. 2. The source (Sc) and microphone positions (m2 etc) tested

What we found

Reverberation time

Reverberation is a key acoustic character of a place. Created by reflections from the stones, it makes the sound linger a little before dying away. It also amplifies the sound. The reverberation time is how it is guaged in architectural acoustics. Our results are shown in Figure 3.

Fig 3. Average reverberation time (T30) for different configurations tested with 95% confidence limits.

Once most of the stones have been removed, and only the 5 sarsen trilithons are left, then the reverberation time drops considerably to around 0.4 seconds. This is as expected because we’ve removed many of the stones the sound reflects from. Figure 4 shows three impulse responses for three configurations. The lack of reflections later on in the bottom plot is clearly visible.

Fig. 4. Impulse responses for 2000 Hz octave band for three configurations for microphone position m23. Sources of some reflections annotated.

One surprise for me was that the reverberation time wasn’t even lower with just the 5 sarsen trilithons in a horseshoe. But looking at the plan and thinking of simple sound paths, there are always potential reflection paths between source and receiver even if only the 5 trilithons are left. (Imagine playing a game of snooker and how the balls would bounce off the stones!)


What about the differences between the other configurations? Looking at the reverberation times in Figure 3 there appears to be significant differences. For example, the reverberation time for the full model (line A) is clearly lower than the other three cases (B, C & D) at 500 Hz. The key to understanding this, is to note that the reverberation comes from sound bouncing back and forth horizontally, because any sound going vertically is lost forever into the air. The bluestones in the full model (configuration A) reflect some of the sound vertically into the air, and this energy is then lost from the circle and the reverberation time is lowered. When the bluestones are removed (B&C), these vertical reflections are reduced contributing to a longer reverberation time. The difference at 500 Hz is statistically significant but perceptually it’s a small difference [2].

This scattering of sound vertically by the bluestones is frequency dependent because of the different relative size of the sound wave and the stones. The bluestones are 0.85m to 1.85m in height; the sound wavelength varies from 4m to 6cm over the frequency range we’re analysing [4].

  • The lowest frequency tested (125 Hz), has the very long wavelength (up to 4m). Here the sound waves are larger than the bluestones and so they simply bend around the stones.
  • For 250 and 500 Hz, the wavelengths are of similar size to the stones and you get more vertical reflections (created via diffraction) and reduced reverberation time.
  • At higher frequencies (1000Hz and above), the wavelengths are smaller than the stones and so the vertical reflections are reduced and therefore the reverberation time is less influenced by the presence of the bluestones.


The effect of the outer sarsen circle lintels is slightly more complex. The comparison to make in Figure 3 is between C (with lintels) and D (no lintels).

There’s a significant difference at the lowest frequency (125 Hz). The sarsen uprights are a similar size as the wavelength in this octave band. When the lintels are added, the effective length of the sarsen uprights increase and there is more vertical scattering in the 125 Hz octave band. This lowers the reverberation time [5].

The other significant differences happen at 2000 Hz and 4000 Hz, where the addition of lintels increases the reverberation time a little. The lintels raise the height of the enclosing stone circle, providing more reflecting area. This would be expected to increase the reverberation time [6].


The reflections from the stones amplify the voice, and as I wrote in a previous blog this can aid speech communication. Figure 5 shows how much sound is amplified by reflections in the different configurations for each microphone position.

Mircophone position m32 leaps out as being different from the others. This is a position where the line-of-sight between source and receiver is blocked by one of the large inner trilithons (see Figure 1 or 2). Configuration E (just the 5 sarsen trilithons in a horseshoe) is 5 dB quieter than the other 4 configurations. Without supporting reflections from the outer sarsen circle, the sound at microphone m32 is much quieter for configuration E.

Fig 5. Amplification created by stone reflections (and ground reflection) for each configuration and microphone position.

The other microphone positions don’t show any clear story. I’m surprised by how configuration E (5 sarsen trilithons in a horseshoe) is not much quieter. On average it’s 2 dB quieter, but that isn’t a large change. This happens because the horseshoe shape of the trilithons means that all microphone positions get some beneficial reflections even for configurtion E. You can see this in the impulse responses of Figure 4, where I’ve circled the reflection from the inner trilithon on the bottom plot.


It has been suggested that the circle of outer sarsens might focus sound and create an echo. But as I examined in a previous blog, for the full model the bluestones and trilithon horseshoe get in the way and this focus isn’t audible for speech. I’ve been listening to the different model configurations to see what that reveals about echoes. The auralisation below is for a balloon burst through the five configurations I measured. I used a balloon pop, because it is an impulsive sound that will make it easier to hear any echoes. The sound file starts with the full model (A), then goes through B,C and D ending with the 5 sarsen trilithon horseshoe (E). This is for microphone m23.

There are two effects I notice:

  • The less reverberant and deader last balloon burst is for the 5 sarsen trilithon horseshoe. The reduction in reverberation shown in Figure 3 is cleary audible.
  • The balloon burst isn’t a single bang, but there is a double impulse. It’s hard to hear for the first balloon burst (full model) but clearer for the other 4 cases. In the full model, masking reflections from the inner bluestones makes the subtle echo hard to hear. The subtle echo comes from one of the trilithons in the horseshoe (see bottom impulse response in Figure 4).

Half the microphone measurements have this subtle double bang, but for the half the only audible effect is the reduced reverberance for the last configuration E.

But even when you get a subtle double bang, it would have been extremely hard to hear with speech, as this auralisation demonstrates:

If our ancestors were making percussive sounds, say for example tapping the stones, then they would be able to perceive some very subtle echoes in some places. These would have been most audible when there were fewer bluestones.


What does this mean for the acoustics of the different construction phases given by Darvill et al?

2620-2480 BC

From a practical perspective, it seems likely that the sarsen trilithon horseshoe was built first [1]. At this point there would have been reflections from the large uprights. Close-by the effects of these reflections is subtle, mostly a little amplification of the voice. Further away than we measured, the reflections would have arrived later and some echoes would have been audible. If a double bluestone circle was added, then this would have weakened any echo effects.

The addition of the outer sarsen circle would have been the most significant addition to the monument. It would have increased the reverberation time by a clearly noticeable amount and provided additional amplification of the voice. Speech communication would have been easier within the circle, but talking from inside to outside the circle would have been made harder.

The lintels of the outer sarsen circle only make very subtle changes to the acoustic. And because of the time required to put them in place, the changes to the sound would have been too small to be detectable by our ancestors.

2480–2280 BC and 2280–2020 BC

The various rearrangements of the bluestones only make very subtle changes to acoustic. The effects we report here are only audible if you have sound recordings and can juxtaposition before and after examples to listen to. Given the time required to reposition all the stones, there is no possibility of our ancestors detecting the difference. The biggest difference you can get with the bluestones is to remove them entirely from the monument (and even then the change is very small). However, it seems unlikely to me that during the rearranging of the bluestones our ancestors would have started by removing them all from within the circle.


[1] Darvill, T., 2016. A Research Framework for the Stonehenge, Avebury and Associated Sites World Heritage Site. Research Activity in the Stonehenge Landscape 2005-2012.

[2] The significance of the changes in reverberation time in Figure 3 were tested using a one-way ANOVA with repeated measures. For example, at 500 Hz there was a significant main effect for ‘Configuration’ on the reverberation time (F(4, 20) = 148.3, p = <0.001, η2=.97, large effect). Bonferroni post hoc tests showed that reverberation time was significantly lower for case E (mean = 0.38 ± 0.04s) compared to the other four cases.

Comparing the other four cases, removing the bluestones (C) made the biggest change to 500 Hz reverberation time: A, full = (0.65 ± 0.06s) vs C, no bluestones = (0.75 ± 0.06s). This is significant but perceptually it’s a small difference. (The just noticeable difference (JND) is about 0.06s [3]).

[3] Niaounakis, T.I. and Davies, W.J., 2002. Perception of reverberation time in small listening rooms. Journal of the Audio Engineering Society, 50(5), pp.343-350.

[4] Jens Holger Rindel derived a really useful concept for working out how scattering from plane surfaces varies with frequency, see Cox, T.J. and D’Antonio, P., 2016. Acoustic absorbers and diffusers: theory, Design and Application. pp. 375 You work out a cut-off frequency, above which you get stronger specular reflection and below which more vertical scattering reduces the reverberation time.

[5] This is one possible explanation. Using Rindel’s formulation for cut-off frequency (see note 4), I get 110 Hz with lintels and 160 Hz without lintels. Based on average stone-receiver distance (15m) and normal incidence. Another explanation might relate to the air gaps between the sarsen uprights.

[6] The lintels above the gaps between the sarsen uprights would also promote some vertical scattering at these higher frequencies. This would act to reduce the reverberation time. But this doesn’t appear to be the dominant effect.