Aarhus University Seal

Dendrophobia in bonobo comprehension of spoken English

Guest lecture by Rob Truswell, University of Edinburgh, UK

Info about event


Thursday 25 August 2022,  at 11:15 - 12:30


Building 1481-324


Ken Ramshøj Christensen

This talk summarizes my 2017 paper "Dendrophobia in bonobo comprehension of spoken English", comparing the acquisition of hierarchical phrase structure by bonobos and humans.  The empirical core of the paper is a reanalysis of data from Savage-Rumbaugh et al (1993), comparing the responses of a bonobo (Kanzi) and a human infant (Alia) to spoken English requests.  Kanzi generally performs accurately in response to requests like (1-2), but his accuracy dips significantly in response to requests involving NP-coordination, like (3).

1) Put some oil in the tomato.

2) Go get the lighter that's in the bedroom.

3) Give the water and the doggie to Rose.

In one respect, Kanzi's performance in response to requests like (3) is at chance: he is equally likely to give just the water, just the doggie, or both.  Alia showed no such problem in interpreting NP-coordination, with no parallel dip in performance.  This is therefore a species-specific, construction-specific deficit.

I demonstrate that the successful interpretation of NP-coordination requires a representation of sentences like (3) involving hierarchical constituency structure: minimally, "water and doggie" must be interpreted as a complex object of "give".  In contrast, most other requests directed at Kanzi can be successfully interpreted using nothing more than the lexical semantics of content words ("put", "oil", "tomato"), statements about linear order (given "put N1 N2", N1 is the THEME and N2 is the GOAL), and heuristics about plausible actions (the GOAL of "give" is usually animate).  So Kanzi's poor performance in response to utterances like (3) would be explained if he did not have access to hierarchical representations of sentence structure.

This is broadly consistent with Fitch's (2014) "dendrophilia hypothesis", which states that humans, unlike most other species, have a "propensity ... to infer tree structures from sequential data".  However, comparison with human acquisition studies suggests that this propensity develops gradually.  Using a preferential looking paradigm, Gertner & Fisher (2012) found that 21-month-old infants treat sentences like (4) as describing a transitive action (a boy as an AGENT and a girl as a PATIENT) rather than a coordinated intransitive action carried out by a boy and a girl.

4) The boy and the girl are gorping!

This suggests that 21-month-olds, like Kanzi, do not parse "the boy and the girl" as a coordinated NP.  In turn, this implies that the dendrophile propensity identified by Fitch is one that develops over time.  Comparatively, the main difference with Kanzi is he never acquired such hierarchical representations, whereas humans eventually do.  In other words, in this respect, Kanzi is a dendrophobe.

Participation is free and open to everybody.

The event is financed by SCC’s research programme for language, linguistics, and cognition