3. 3
ABSTRACT:
This study examines animal sounds and their onomatopoeic equivalents from a phonetic point
of view that may shed light on the process of human perception of animal sounds as well as
examining possible factors that affect perception. Data was gathered from video clips of
animals and speakers producing their languages’ onomatopoeic words and analyzed on Praat.
In addition, syntheses were created to see how closely the sounds could be reproduced based
on vowel formants. I hypothesized that mammals sounds were most easily replicated by human
languages and were similar cross-linguistically, but found that while two of the three cross-
linguistically similar onomatopoeia were of mammals, other mammals showed various degrees
of dissimilarity both cross-linguistically and with regards to the actual animal call. Because of
the nature of the methodology and time constraints, this study remains inconclusive but sheds
some light and raises questions on the role of physical relatedness and onomatopoeic sounds.
INTRODUCTION
Each culture has onomatopoeia and a canonical way of representing animal sounds. Sometimes
these sounds are similar as in cows going “moo” in English and “mo” in Japanese, while at other
times they may differ radically such as roosters going “cock-a-doodle-doo” in English and
“wowowo” in Mandarin. In this project, I intend to examine the variety of animal sounds and
compare their onomatopoeic sounds cross-linguistically. Based on the formant analyses, I will
also examine whether certain features or animal classes may have an effect on onomatopoeia
matching actual animal sounds.
4. 4
METHOD
This project aims to analyze animal sounds and as well as their onomatopoeic counterparts in
various languages to determine if any acoustic similarity is present and to investigate which
factors may lead to similarities between animal utterances and onomatopoeia as well as
onomatopoeia cross-linguistically.
Animal sounds were selected based on perceived commonality and availability. Sites including
animal onomatopoeia were consulted to determine which sounds seemed most common. Next,
animal sound clips were gathered from online sources, mainly from the video-sharing site
YouTube.com. Animal sounds without significant background noise were the most optimal
sources for analysis, but suitable clips were difficult to acquire. In the end, eight animals were
selected: dogs, cats, pigs, cows, roosters, ducks, frogs, and bees.
The foreign language onomatopoeia tokens were selected by Internet availability for each of the
eight selected animal sounds. I chose to use only those onomatopoeia that I could find sound
files for, despite the large amount of written information on foreign language onomatopoeia,
which was used as a reference (Abbott, 2004). This was to ensure that phonetic similarities could
be more accurately determined, as written words do not always accurately indicate the actual
phonetic production. Also, sounds that were not in the language’s phonology were removed; thus
pulmonic ingressives and clicks, which existed in some videos, were assumed to not be a lexical
item, but a direct imitation of the animal sound. The total list of languages ended up as: Arabic,
Bengali, Dutch, English, French, German, Hindi, Hokkien, Indonesian, Italian, Japanese,
Korean, Mandarin, Romanian, Russian, Spanish, Tagalog, Turkish.
5. 5
These sounds were then transcribed into narrow IPA and compared cross linguistically to find
patterns. If applicable, different onomatopoeic paradigms were constructed to categorized
different onomatopoeia. These were plotted on a map and geo-linguistic patterns were analyzed.
Finally, rudimentary syntheses based on formants in the animal clips were created for some
animal sounds in order to discover whether or not certain parts of the onomatopoeic sounds
could have been based on the formants in an animal’s utterance as well as to see any unexpected
formant or other articulatory difference reflected in the spectrograms of animal sounds.
Results
Chart
1.
dog
cat
pig
cow
rooster
duck
bee
frog
tagalog
aw
aw
oink
mo
tik-‐tilawk
kwak
kokak
indonesian
kʊk
kʊk
mɛɐ̃ŋ
ŋok
mü
kukuruyuk
kwek
ŋuŋ
krok
korean
mʌŋ
jaw
kɯɭɯ
ɯmme
kokioː
kwɛk
kʰɛkɯɭ
japanese
waɴ
nja
/njaɴ
bɯçi
moː
kokekokko
kuakku
bɯn
keɭo
english
wʊf
mjaʊ
ojŋk
muː
kʰɒkʰədudl̩duː
kʰwæk
bʌz
rɪbɪt
german
vʊf
mjaw
gʁʊnts
muː
kikeriki
kʰwak
zʊm
kʰwak
dutch
wɜf
mjaw
knor
bubu
kʉkelekʉ
italian
baw
mjaw
muu
kikiriki
kwak
zz
kra
kra
french
waf
mjaw
gʁwã
mø
kokoriko
bzz
spanish
ɣʷaw
mjaw
muu
kikiriki
bzz
romania
hɐm
mjaw
grʊf
mou
russia
gav
mjaw
xrju
mu
kukareku
india
baw
waw
mjaw
bawwi
klʌk
klu
klʌk
bengali
ɡʰew
mew
hamba
molok
arabic
haw
mjaw
bak
bvakgir
zzt
turkish
haw
haw
mjaw
møː
ʉ
ʉrʉ
ʉ
ʉː
mandarin
waŋ
waŋ
mjaw
mow
ɔ
ɔ
ɔ
taiwanese
wɤŋ
wɤŋ
mjaw
mow
kokokoo
bʑʑ
6. 6
Dog
+cont[ ]( )
V
+back
!
"
#
$
%
& +cont[ ] eg. [waŋ]
While dog sounds could all be analyzed into the structure above, several subdivisions seem to
exist which have greater within-group similarities than between-group similarities, especially
given the geo-linguistic ties within these groups (see appendix A1).
1. West Germanic + French:
LAB
+cont
!
"
#
$
%
&
V
+back
!
"
#
$
%
& f[ ] eg. [wʊf]
2. East Asian Sprachbund:
LAB
+cont
!
"
#
$
%
&
V
+back
!
"
#
$
%
&
DOR
+nasal
!
"
#
$
%
& eg. [wɤŋ]
3. Scattered languages: (C) [aʊ] eg. [baʊ]
Three dog barks are shown in the spectrogram below. F2 and F3 generally seem to fall while F1
exhibits a slight rising and falling tendency, most clearly in the third bark. In addition, the
intensity seems to fall just before F1 falls in frequency, with the darkest energy bands happening
in the front. Without any noticeable difference in the background grey bands, it seems this most
closely resembles a low to high diphthong, something like /aw/. Indeed, the synthesis based on
these formants produced a low to high rising diphthong [aɰ]. Thus, an onset-less category 3 type
sound seems most similar.
7. 7
Pigs
DOR
+voice
!
"
#
$
%
& rhotic[ ]
LAB
−cons
!
"
#
$
%
&(C)(C)(C)
(see appendix A2)
eg. [gʁwã]
Pig sounds varied significantly. In addition, the spectrogram contained various phenomena other
mammals did not exhibit such as creakiness, shown by lines in the energy, and a static F1 that
mirrored closely the F2. Removing F1 from the synthesis created a sound similar to [əәi]. No
nasals can be clearly seen, nor final or initial consonant-like sounds. Also interesting is that the
higher formants seem to have successively later voice onset times. This analysis remains
inconclusive although the vowel indicated by formants and the synthesis points to an [oiŋk].
Rooster
[[k]V]σ [[k]V]σ [[liquid]V]σ [[k]V]σ (see appendix A3) eg. [kokoriko]
The rooster’s sound seems to have an overall syllabic shape cross-linguistically with three or
four syllables. This is reflected in the dips in intensity at specific parts. These dips in intensity
co-occur with slight dips in frequency. The large blank area with massively dipping intensity in
the beginning also may be interpreted as a stop, which lends support to the “cock-a-doodle-doo”
of English onomatopoeia. Another point of interest is that so many languages have a k as onsets
8. 8
for each of these syllables, but given the geolinguistic ties, it may be loan word influence.
Another interesting finding is in the South Asian and Southeast Asian regions, several languages
represented the rooster’s call with a closed syllable with coda [k], despite the last part of the
spectrogram showing the clearest signs of pure vowel formants. The synthesis based on vowel
frequencies of F1 and F2 yielded a sound akin to [ɐ].
Duck
k[ ] w[ ]
V
−high
"
#
$
%
&
' k[ ] eg. [kwak]
The duck sounds exhibited an amazing degree of similarity with the majority of difference
existing solely in vowel height from [e] to [a] and aspiration of the velar stops. The below
spectrogram demonstrates this, showing a clear F2 and F3 transitioning from a single point into
two clear formants, along with a slight frication band of energy in higher frequencies. F1 also
appears as an arced band of faint energy. While a final velar pinch is not clearly seen, the
formants move in that position. Also, there is not a frication, but as in human speech, and also
9. 9
because of the information-carrying capabilities of formant transitions, the sound is likely to
cause humans to perceive a [k]. Why there is a [w] glide is not as clear.
Frog
Western European Sprachbund:
DOR
−son
−cont
"
#
$
$
$
%
&
'
'
'
+son
−nasal
"
#
$
%
&
' V[ ]
DOR
−son
−cont
"
#
$
$
$
%
&
'
'
'
(see appendix A4)
eg. [kroʊk]
Another interesting phenomenon is the frog croak, which cross linguistically is commonly
treated as another [k] sound. Here, we also see a velar pinch in both ends of the utterance as well
as lines showing a creaky voice. However, an [r] or [w] is not as visible since F2 and F3 stay
stable after transitions. It is possible that one of the other sounds is more accurate.
Bee
LAB
−cont
"
#
$
%
&
'V z[ ] eg. [bʌz]
Bee sounds are rather uniform with a voiced coronal fricative in all onomatopoeia found. Despite
the sound being produced not orally, but by the beating wings, the source is still turbulence and
therefore acoustically is similar. The initial [b] in many onomatopoeia could be the interpretation
10. 10
of the acceleration of the wings before full speed, as a lower frequency is characteristic of labials
and therefore it may be perceived as a labial.
Cat
+nasal[ ] j[ ] V[ ]
−cons
+round
"
#
$
%
&
'
(see appendix A5)
eg. [mjaʊ]
Cats had a very similar onomatopoeic representation throughout the world, with the vast majority
of languages all phonetically representating their cries as [mjaʊ]. If we look at the spectrogram,
it appears that the diphthong [aʊ] is quite clear in the formant movement, and the initial rising
band of energy may hint at a perceivable [j]. The nasal is less clear, although there is a faint band
of energy that is both visible and being picked up that may be nasal damping effects and thus
hint at a nasal onset.
11. 11
Cows
m[ ]
V
+back
+round
+long
!
"
#
#
#
#
$
%
&
&
&
&
eg. [mu]
Cows similarly had near universal onomatopoeia. The spectrogram too showed most clearly the
possible reasons for this similarity. Clear formants gave a pure long vowel. The synthesis gave a
vowel similar to [ɒ], but I suspect that the sound is typically is given an [u] or [o] sound in
onomatopoeia because of the nasality seen in the beginning of the call heard throughout the
vowel. The nasal qualities may have effects on perception.
CONCLUSION
&
DISCUSSION
In conclusion, I found that the closest phonetic representations of animal sounds did not line up
as closely with biological similarity. Cats and cows, both mammals, had near universally similar
onomatopoeia which closely paralleled the spectrogram formant patterns. However, ducks,
which are not mammals, produced sounds that were similarly near identical with spectrogram
readings that did contain recognizable features. Less universal were dogs, roosters, and bees,
none of which were from the same animal family but showed a medium amount of phonetic
12. 12
features cross-linguistically although these were not necessarily as traceable to the actual
spectrograms as the animals with the most similar onomatopoeia. Lastly, pigs and frogs showed
a high degree of variability with only one or two similar features: the [k] sounds with frogs.
The presence of mammals in various tiers of onomatopoeic similarities does not lend support to
genetic and physiological similarity as a factor. Still, as this was a phonetics study, it could be
that the assumption that similar genetics would mean similar physiology would need more
justification as it could be that certain animals even within the same family could have evolved
apart in the vocal tract area, or converged despite being in different evolutionary trees. Indeed,
this may be the reason why some of the syntheses had acoustic productions that differed from the
animal sound expected or suggested by the majority of surveyed languages.
A further area of study is whether or not certain sound patterns in animal calls are more easily
perceived. For example, both cows and cats share nasality in their onomatopoeia and this nasality
can be seen somewhat in the spectrograms, while pig and frog calls both exhibit creakiness.
13. 13
REFERENCES
Abbott, D. (2004). Animal Sounds. Retrieved from
http://www.eleceng.adelaide.edu.au/personal/dabbott/animal.html
ESL Web. (ND). Hear What?: Animal Sounds in different languages. Retrieved from
http://www.esl-languages.com/en/animal-sounds.htm
Funnysexy0624. (2, Nov, 2010). My Korean Boyfriend: Korean animal sounds. Retrieved from
http://www.youtube.com/watch?v=IfN6tYHJWZU
Immsl. (21, Aug, 2009). A Bee Buzz At My Sitting Room. (Very Noisy Buzz). Retrieved from
http://www.youtube.com/watch?v=LsDM-ktYLsc
Keirmorse. (1, May, 2007). Pacific Tree Frog. Retrieved from
http://www.youtube.com/watch?v=XcFKQKjv0-o
Kidslearningvideo.(30, Aug, 2010). Farm Animal Sounds. Retrieved from
http://www.youtube.com/watch?v=vuiwA4Ne_pU
Properniceinnit. (ND). Bow Wow Meow – Animal Sounds in Different Languages. Retrieved
from http://vimeo.com/25215616
WoodGirl14. (20, Oct, 2009). Animal sounds in Japanese, Indonesian, German, Italian.
Retrieved from http://www.youtube.com/watch?v=CEpudI87Pew