AnimalCommunication - Newcastle University · Why Look At Animal Communication? Kanzi Songbirds Humpback Whales References Truswell,R.2017. Dendrophobiainbonobocomprehensionof spokenEnglish.

Animal Communication

Rachael Bailes

SEL1008

Tuesday 27th November 2018

Why Look At Animal Communication? Kanzi Songbirds Humpback Whales

OutlineWhy Look At Animal Communication?

HomologyAnalogyDesign Features of Language

KanziSymbol useStructure sensitivity or linear order?

SongbirdsVocal learning and cultural transmissionZebra FinchesSong Sparrows

Humpback WhalesHumpback Whale Song contact

2 / 47


Why Look At Animal Communication?

• Animals communicate with each other using (sometimesvery complex) signalling systems

• The extent to which animal communication systems aresimilar or different to human language can be informative

• It helps us probe how communication works in the naturalworld, and when/how human language may have evolved

• Two important concepts to help navigate this arehomology and analogy.

3 / 47


Homology

• Homologous traits are features of different organisms thatare similar because they were inherited from a commonancestor that also had such a trait.

• Example: Birds, bats, mice, and crocodiles all have fourlimbs (tetrapods), whereas sharks and goldfish do not

• This is because the last common ancestor (LCA) oftetrapods evolved four limbs, and the descendants inheritedthis trait

• The presence of four limbs in birds is homologous withthe four limbs of mice; both instances share anevolutionary origin

4 / 47


Homology

Source: evolution.berkeley.edu/evolibrary/ 5 / 47


Homology• When we look for homologous traits to do with language,we are primarily studying our nearest extant relatives,usually chimpanzees (Pan Troglodytes) and bonobos (Panpaniscus) - though sometimes other primates too

• LCA of Homo sapiens and Pan was roughly 8 million yearsago (though the split likely spanned millions of years)

6 / 47


Analogy• Analogies, by contrast to homologies, have distinctevolutionary origins

• Example: Bat wings and bird wings look superficiallysimilar, because they have both evolved to solve theproblem of natural flight

• But bats are not more closely related to birds than theyare to mice; each type of wing has evolved independently

Source: evolution.berkeley.edu/evolibrary/

7 / 47


Analogy

• Bat wings are analogous to bird wings - they look similarbecause they have been shaped by similar selectivepressures, but have followed independent trajectories

• Analogies can provide a good model for how traits work,and help explain why they take the form they do

Source: evolution.berkeley.edu/evolibrary/

8 / 47


Design Features of Language

• We have good reason to believe that humans are unique intheir linguistic ability; no other animal on earth haslanguage

• We also know that language is made up of manyconstituent abilities and features

• Syntax, symbolic ability, vocal learning, inferential abilities,cultural transmission, etc...

• To what extent to other animal communication systemsshare these design features? Which parts of language areunique, and which parts are complex versions of what otheranimals have?

• Which features do we share with our closest relatives?Which features have analogues elsewhere in nature?

9 / 47



• Hockett (1960) gave a formalised list of features thatcharacterise human language, and set it apart from animalcommunication

• Some of these design features of language are featured inother animal communication systems, but only languagehas all of them

10 / 47



11 / 47


Chimps can use symbols• Various attempts to teach primates sign language (e.g.chimpanzees Washoe and Nim Chimpsky, Koko thegorilla), with varying success. Project Nim is on Netflix -give it a watch.

• “Give orange me give eat orange me eat orange give me eatorange give me you”

• Kanzi achieved high competence of English-likecommunication system (spoken, lexigrams) without explicitinstruction & acquired a lexicon of several hundred items(Savage-Rumbaugh et al, 1993)

• Kanzi and Alia (human 18mo infant) were tested over 9months on 660 spoken instructions to deal with familiarobjects in novel and unpredictable ways. The instructionsand responses were recorded and coded for success in acorpus.

12 / 47


Kanzi

• Savage-Rumbaugh et al (1993) reported an overall accuracyof 71.5% for Kanzi - slightly higher than Alia’s 66.6%

• Of particular interest is Kanzi’s ability to comprehendinstructions that involve a direct and indirect object, suchas:

• https://www.youtube.com/watch?v=2Dhc2zePJFE

13 / 47


Kanzi: structure-sensitivity?

V’

V

put

XP

NP

D

the

N

tomato

X’

X

(LOC)

PP

P

in

NP

D

the

N

oil

14 / 47


Kanzi: structure-sensitivity? Not so fast

• Truswell (2017) points out that a simpler strategy fordealing with utterances like put the oil in the tomato / putthe tomato in the oil is to use the linear order:

• first noun is theme, second noun is goal

15 / 47


Kanzi: NP-coordination (Truswell, 2017)• NP coordination: creates a noun phrase of the form NP1and NP2

• Fetch [[the ball] and [the rock]]• In order to get the right response, Kanzi has to representtwo NPs as inhabiting one constituent, whose meaning isdetermined jointly by both component NPs

• In other words, ‘fetching’ has to happen to the objectdenoted by each coordinated NP

• Three likely outcomes: ignore NP1, ignore NP2, fetch both.(ignore both is also possible, but the worst possiblestrategy in any event & least likely)

• A null hypothesis of Kanzi’s structure sensitivity: Whenpresented with coordinated NPs, Kanzi is right 33.3% ofthe time (ie. chance performance).

16 / 47


Kanzi: NP-coordination (Truswell, 2017)

• Kanzi’s accuracy on trials with coordinated NPs was 22.2%- significantly worse than his overall score of 71.5%

• Worse than chance performance - structure-sensitiveinterpretations were dispreferred.

• Alia’s accuracy on coordinated NP trials was 68.4% - inline with her accuracy across the whole corpus.

• Kanzi’s performance dips dramatically on precisely thetrials that require understanding of hierarchical structure,rather than simpler strategies like linear order or ‘semanticsoup’.

• Although Kanzi was more successful than Nim, and showsimpressive comprehension of novel strings, he still lacks theability to deal with constituents in a hierarchical structure

17 / 47


Other primates can’t handle recursion

• Chimps, which are very similar to humans genetically(about 98%) and in some cognitive respects, cannot learnhuman language like human children.

• No animals clearly show recursion in their communicationsystems, or are able to learn recursive structures.

(1) [IP Rachael said [CP that [IP she welcomed our newalien overlords ]]

(2) [IP Rachael said [CP that [IP [IP she welcomed ournew alien overlords ] and [IP they approved ] ] ]

18 / 47


What else could we look for?

• Human language produces a set of sounds (or gestures)that is highly structured, i.e. it has a syntax.

• Human language has to be learned, in the sense thatchildren need to match an abstract model pattern (agrammar) that they hear (or see).

• They have to do more than just memorising strings.• The structure seems to be related to the ability to do thistype of learning (grammar induction, as opposed totrial-and-error learning).

19 / 47


Language Transmission / Acquisition

Adults’ I-Language −→ Adults’ E-Language

Children’s I-Language −→ Children’s E-Language

20 / 47


Bird Song

• For songbirds, such as Zebra Finches and Song Sparrows,the song is structured, with individual parts to it (songsyllables).

• The syllables are arranged in a certain order (song syntax).• Bird song has to be learned, in the sense that baby birdsneed to match a model pattern that they hear from adults.

• Like human language, the ability to learn song is innate forthese species, though the specific song must be learned.

21 / 47


Zebra Finches

22 / 47


Zebra Finch Song Syllables

(Tchernichovsky et al, 2004)

23 / 47


Zebra Finch Song Syntax

24 / 47


Zebra Finches• Male songbirds learn their song from their fathers, ingeneral (Slater, 2011)

• After hearing the song, the birds produce an imperfectcopy, called subsong.

• The young bird continues to practice, getting closer andcloser to the adult song, until it arrives at a near-perfectcopy of the father’s song.

• Zebra Finches begin to learn song at 25 days afterhatching, and can produce their first subsong at 30 days.

• About two weeks later, they produce identifiable songsyllables.

• By day 90, the finch has learned the order of the syllables,and their song does not change much after that time.

25 / 47


Development of a Syllable

26 / 47


Zebra Finch Song Learning

• Zebra Finches show a critical period for song learning,from day 25-90.

• Similar to the critical period for human language learning(roughly in utero – pre-puberty, around age 9 according toNewport, 2002).

27 / 47


Innate Ability to Learn Song

• If a bird is raised in isolation, it still produces a song, butone without a full song syntax (Slater, 2011).

• Similar to children who are deprived of language exposureduring the critical period.

• If a group of birds are raised in isolation, the firstgeneration produces a simplified song, but followinggenerations learn a full adult song using the simple song asthe model.

• Similar to the emergence of sign languages, or to languagevarieties arising out of language contact situations.

• The last two facts imply that the ability to learn song isinnate.

28 / 47


Zebra Finch Song Development, over time(Tchernichovsky et al, 2004)

29 / 47


Zebra Finch Song Development

30 / 47


Song Sparrows

• Another aspect of human language is sociolinguisticvariation.

• As language changes, variation arises in the population.• The different variants can sometimes take on socialmeanings.

• Does any animal communication do this?• In Zebra Finches we saw learning, which is necessary for

change to occur.• Song Sparrows take this one step further, and assign social

meaning to the changes in the song.

31 / 47


Song Sparrow

32 / 47


Song Sparrows

• Male song sparrows learn their father’s song, much likeZebra Finches.

• But then they go out, claim some area, and learn some ofthe songs of their male neighbors (at least during their firstSpring season (Liu’ and Kroodsma, 2006)).

33 / 47


Song Sparrows

• They can choose which of the songs in their repertoire tosing in response to their neighbors songs, and these havedifferent meanings (Beecher & Campbell, 2005).1. If a bird sings a song, and its neighbor sings the same song

back, that is very aggressive (territorial challenge).2. If a bird sings a song, and its neighbor sings a different song

back, but one that they share in their repertoires, that isless aggressive.

3. If a bird sings a song, and its neighbor sings a different songback that they don’t share, that is the least aggressive.

34 / 47


Graveyard Study(Liu’ & Kroodsma, 2006)

35 / 47


Repertoires of Two Sparrows – 3 shared songs out of 9(Beecher & Campbell, 2005)

36 / 47


Brain Area Comparison(Jarvis, 2004)

37 / 47


Humpack Whale Song

• Males sing a song, with different parts to it. We do notknow why.

• All the males in a pod sing the same song, but the songchanges a little bit over the singing season and theytransmit changes in the song to each other.

• Since they are capable of spreading changes, they arecapable of vocal learning (Janik, 2011)

• A song can change completely over a period of 12 years(faster than human language, if whales live around 50 yearsor more).

38 / 47


Humpback Whale

39 / 47


Humpack Whale Social Situation

• As with human language contact, the social situationmatters with whale song. (We don’t know why, but thesong seems to have social meaning.)

• This makes sense, since humpback whales are socialanimals, like humans (much more so than songbirds).

• The humpback population off the east coast of Australiawas invaded by males from the west coast.

• The song of the east coast whales changed completely tothe west coast song in 2 years (not the usual 12 forcomplete song change).

• Like humans, song contact can create more radical changesin the song of a population than internal change.

40 / 47


Conclusions

• No animals have been discovered to have communicationsystems like human language, in terms of its complexity ofstructure (especially recursion).

• Vocal learning and transmission evolved in humans, butnot in any other primates.

• It is not a feature of the primate family.• You can’t teach a chimp human language.

41 / 47


Conclusions

• However, vocal learning (of a grammar) has evolvedindependently in other species (some cetacean species andsome bird species).

• Note that vocal learning is not dependent on intelligence• Rather, it seems to require connectivity and coordination of

auditory and motor areas of the brain. Humans and othervocal learners have this. Other primates not so much.

• Other features of the signal system, such as syntactic rules,variation & change, critical periods of learning, all seem tobe related to this modality in particular

• For more on Zebra Finches, watch the videos here!http://ofer.sci.ccny.cuny.edu/

42 / 47


Bonus slide: Vervet Monkey Alarm Calls(Seyfarth, Cheney & Marler, 1980)

43 / 47


Bonus slide: Vervet Monkey Alarm Calls

Something to think about until next week:• A given vervet call, or other animal signal, ‘means’something (eagle, leopard, snake)

• How do we know this? What does ‘mean’ mean here?• How do we think vervets derive this ‘meaning’?

44 / 47


ReferencesBeecher, M.D., and S.E. Campbell. 2005. The role of unsharedsongs in singing interactions between neighbouring songsparrows. Animal behaviour 70 :1297–1304.Bolhuis, J.J., K. Okanoya, and C. Scharff. 2010. Twitterevolution: converging mechanisms in birdsong and humanspeech. Nature Reviews Neuroscience 11 :747–759.Deregnaucourt, S., P.P. Mitra, O. Fehér, K.K. Maul, T.J. Lints,and O. Tchernichovski. 2004. Song development: In search ofthe error-signal. Annals of the New York Academy of Sciences1016 :364– 376.Hobaiter, C. & Byrne, R. 2011. The gestural repertoire of thewild chimpanzee. Animal Cognition 14 :745-767.Hockett, C. 1960. The Origin of Speech. Scientific American203 : 88-96.

45 / 47


ReferencesTruswell, R. 2017. Dendrophobia in bonobo comprehension ofspoken English. Mind and Language 32, 4: 395-415.Janik, Vincent M. 2011. Vocal communication in cetaceans. InThe Oxford handbook of language evolution, ed. MaggieTallerman and Kathleen R. Gibson. Oxford: Oxford UniversityPress.Jarvis, Erich D. 2004. Learned birdsong and the neurobiologyof human language. Annals of the New York Academy ofSciences 1016 :749–777.Liu’, Wan-Chun, and Donald E. Kroodsma. 2006. Songlearning by chipping sparrows: when, where, and from whom.The Condor 108 :509–517.Newport, Elissa L. 2002. Critical periods in languagedevelopment. In Encyclopedia of cognitive science, ed. LynnNadel, 737–740. London: Macmillan.

46 / 47


ReferencesSeyfarth, RM. Cheney, DL. & Marler, P. 1980. Monkeyresponses to Three Different Alarm Calls: Evidence of PredatorClassification and Semantic Communication. Science210(4471): 801-803.Slater, Peter. 2011. Bird song and language. In The Oxfordhandbook of language evolution, ed. Maggie Tallerman andKathleen R. Gibson. Oxford: Oxford University Press.Tchernichovski, O., TJ Lints, S. Deregnaucourt, A. Cimenser,and PP Mitra. 2004. Studying the song development process:rationale and methods. Annals of the New York Academy ofSciences 1016 :348–363.Terrace, Herbert S. 1983. Apes who talk: language orprojection of language by their teachers? In Language inprimates, 19–42. Springer.

47 / 47

AnimalCommunication - Newcastle University · Why Look At Animal Communication? Kanzi Songbirds Humpback Whales References Truswell,R.2017. Dendrophobiainbonobocomprehensionof spokenEnglish.

Documents