[Lingtyp] animacy hierarchy: exceptions based on shape: SUMMARY

David Gil gil at shh.mpg.de
Tue Dec 4 16:14:01 UTC 2018

Dear all,

I would like to thank everybody who responded to my query last week on 
exceptions to the animacy hierarchy based on shape (appended below).In 
the rather lengthy summary that follows, I first explain the reason 
behind the query, and then discuss some of the results that emerged from 
your responses.

The motivation for the query is in joint work by Yeshayahu Shen and 
myself on how we conceptualize and talk about hybrid creatures such as 
centaurs and mermaids — see reference below.One of our experimental 
findings was that when people are asked to describe novel hybrids, they 
tend to do so in accordance with the animacy hierarchy; for example, 
when presented with a composite image of a man and an eagle, they are 
more likely to say "man with an eagle's head and wings" (where the NP is 
headed by a human noun) than "eagle with a man's torso and legs" (where 
the NP is headed by a noun that lower on the animacy hierarchy).However, 
more recent work has led us to modify the above conclusion; it now seems 
to us to be the case that the hierarchy that governs the way we describe 
hybrids is, at least in part, not the familiar ontologically-based 
animacy hierarchy, but rather an alternative "schematological" 
hierarchy, based on shape, which, simplifying considerably, looks 
roughly as follows:

entities with head, torso, two arms, two legs, face with sensory organs, 
lateral symmetry

(e.g. humans, some apes, some robots)

entities with head, torso, appendages, lateral symmetry

(e.g. dogs, birds, cars)


entities with head, torso

(e.g. jellyfish, trees, umbrellas)


Although resembling the familiar animacy hierarchy in that it "begins" 
with humans and gradually peels away the human properties, the above 
schematological hierarchy differs from it in that it is based not on 
"deep" ontological properties but rather on properties pertaining to 
relatively more superficial shape.

To the best of my knowledge, there are no reports in the linguistic 
literature of grammatical patterns being governed by the above 
schematological hierarchy; as far as I have been able to ascertain, all 
patterns argued to be sensitive to animacy seem to involve the familiar 
ontologically-based hierarchy and not have anything to do with shape.The 
purpose of this query was to see if anybody might be familiar with 
exceptions to the animacy hierarchy possibly motivated by the 
schematological hierarchy; however, so far no such cases have been offered.

My hope was that some attested grammatical patterns might provide at 
least some additional support for the schematological hiertarchy.  
However, if indeed no such case can be found, then our study of hybrids 
would seem to offer a rather curious instance of a grammatical hierarchy 
being revealed solely by a specific experimental task involving a rather 
esoteric stimulus set — which is a state of affairs I wouldn't quite 
know what to make of.

Some comments on selected points raised in the discussion follow below 
(with apologies to those whose contributions I didn't mention).

There was some discussion of the fact that in Japanese, robots can 
cooccur with either of two existence verbs, /iru/, which selects for 
animates, and /aru/, which selects for inanimates, and there was 
disagreement amongst speakers of Japanese with regard to the crucial 
case of whether statues of people, ontologically inanimate but 
schematologically humanoid, could cooccur with /iru/: while some 
speakers rejected the possibility, others had a more nuanced 
take.Yukinori Kimoto, in an offline communication, wrote that

"Japanese linguists seem to agree that the distinction is made according 
to whether the referent has an internal potential to move autonomously. 
So you can express "there is a ship (moving)" with "iru".The judgement 
is ultimately influenced by the speaker's own construal, but robots can 
basically move by themselves, so "iru" is acceptable, (but "iru" sounds 
strange for industrial robots, which do not move, fixed to one place, 
but a rumba moving in your room can take "iru").Yes, statues might take 
"iru", but it is just marginally acceptable. You can say "dozo ga iru" 
(statue NOM exist), in a "spooky" context, like, if you unexpectedly 
find a statue. [...] You can use "iru" with "dozo" (statue) like in the 
"spooky" context where you are walking in the campus at night, and 
surprisingly you find something like a human, which turns out to be a 

Thus, the distinction seems to revolve around the ability to move 
autonomously, while shape is irrelevant except as an indirect predictor 
of whether the object in question possesses such an ability.

More generally, I agree with Östen Dahl, who wrote that

  "The reason why we might want to regard a statue as animate is not 
primarily that it is shaped like a human being but that we either take 
it to represent a human or to be capable of acting and perceiving."

Or as Sebastian Nordhoff put it

"It would seem that "(potentially) having a soul" would be more 
important than geometrical "humanoid" shape."

So the bottom line is that the schematological hierarchy proposed above 
seems to have little if any grammatical manifestations in human languages.

On a somewhat different note, Martin Haspelmath argued that binary 
contrasts such as Japanese /iru/aru/ are more appropriately analyzed as 
reflecting nominal classification rather than true hierarchical 
organization.My response to this would be to invoke Martin's own 
distinction between language-specific descriptive categories and 
cross-linguistic comparative concepts.Internally to Japanese, I would 
agree that the /iru/aru /distinction does not constitute evidence for a 
hierarchy; in order to have a language-specific hierarchy, the language 
would have to make reference to three or more distinct values, such as, 
for example, in Navajo, where an animacy hierarchy governs the choice of 
clausal voice.However, as typologists, it does in fact makes sense to 
posit an animacy hierarchy as a comparative category, reflecting the 
different cutoff points that various languages make, even if some of the 
relevant categories happen, within individual languages, to be binary, 
such as is the case for the /iru/aru /distinction in Japanese.

Several respondents mentioned numeral (and other) classifiers, and 
offered interesting insights as to how these work.However, such 
classifiers fail to fit the bill for the following reason.To the extent 
that they make reference to properties involving shape, they refer to 
indivisible atomic shapes such as "long thin object", "small compact 
object" and the like — not to the kind of complex shapes that form the 
basis of the schematological hierarchy above.Perhaps because of their 
atomic nature, classifiers do not seem to organize themselves in terms 
resembling an animacy (or schematological) hierarchy, in which types are 
arranged in terms of increasing/decreasing complexity.But I'm open to 
counterexamples on this, and Daniel Ross has offered some suggestive 
leads on gender resolution in Bantu needing further exploration.

Related to this, Yukinori Kimoto wrote that

"Japanese classifiers do have a hierarchical structure, but the 
higher-order distinction is between human, non-human animate, and 
inanimate, so again it follows animacy. Yo Matsumoto in an article says 
(translation mine from Japanese):"The most basic semantic distinction in 
Japanese classifier system is animacy. Classifiers are divided into 
those used for animate things and those for inanimate. The first is 
further divided into those used for humans and those for non-human 
animals. Each of the three classes has the superordinate classifier: 
-ri/-nin (for human in general) -hiki (for animals in general), and -tsu 
(inanimate in general)" (parentheses mine)".

However, no evidence is offered for the hierarchic nature of the 
"higher-order" distinction between human, non-human animate, and 
inanimate.(And in any case, even if it turns out that there is 
hierarchical structure, it would seem to follow an ontological rather 
than a shape-based hierarchy.)

Finally, Eitan Grossman offered a plausible frequency-based explanation 
for WHY the schematological hierarchy is not as grammaticalized as the 
ontologically based animacy hierarchy:we talk less about the shapes of 
things than about their deeper ontological properties, and hence shapes 
have less of a chance to undergo grammaticalization.

Thanks to all those who contributed to the discussion:Östen Dahl, Victor 
Friedman, Eitan Grossman, Martin Haspelmath, Joo Ian, Yukinori Kimoto, 
Amina Mettouchi, André Müller, Randy J. LaPolla, Sebastian Nordhoff, 
Paolo Ramat, Jan Rijkhoff, Daniel Ross, Steven Schaufele, Tasaku 
Tsunoda, Kazuha Watanabe.


Shen, Yeshayahu and David Gil (2017) "How Language Influences the Way We 
Categorize Hybrids", in H. Cohen and C. Lefebvre eds, /Handbook of 
Categorization in Cognitive Science/, Second Edition, Elsevier, 
Amsterdam, 1177-1200.

On 27/11/2018 03:27, David Gil wrote:
> I am looking for examples of exceptions to the animacy hierarchy that 
> are motivated by the shape or other spatial configurational properties 
> of the relevant referents.
> The animacy hierarchy is primarily of an ontological nature; shape 
> doesn't usually matter.A slug is animate even though its shape is 
> ill-defined and amorphous, while a stone statue is inanimate even if 
> it represents an identifiable person.
> What would such a shape-based exception to the animacy hierachy look 
> like?In Japanese (according to Wikipedia, I hope this is right), there 
> are two verbs of existence, /iru/ for animates, /aru/ for inanimates, 
> but /robotto/ ('robot') can occur with either of the two: while /iru/ 
> entails "emphasis on its human-like behavior", /aru/ entails "emphasis 
> on its status as a nonliving thing".This description seems to suggest 
> that it's the robot's sentience that is of relevance, not its human 
> shape: presumably, even if the robot assumed the form of a sphere with 
> blinking lights, if its behaviour were sufficiently humanlike it could 
> take /iru/ (speakers of Japanese: is this correct?).On the other hand, 
> I'm guessing that a human-like statue could never take /iru /(is this 
> correct?).So if my factual assumptions about Japanese are correct, the 
> distribution of /iru/ and /aru/ does not offer a shape-based exception 
> to the animacy hierarchy.A bona-fide shape-based exception to the 
> animacy hierarchy would be one in which all human-shaped objects — 
> robots, dolls, statues, whatever — behaved like humans with respect to 
> the relevant grammatical property.Or conversely, a case in which an 
> animate being that somehow managed to assume the form of a typical 
> inanimate object would be treated as inanimate.
> I would like to claim that such shape-based exceptions to the animacy 
> hierarchy simply do not exist, but I am running this past the 
> collective knowledge of LINGTYP members first, to make sure I'm not 
> missing out on anything.
> More generally, it seems to be the case that grammar doesn't really 
> care much about shapes.The closest thing to grammaticalized shape that 
> I can think of is numeral classifiers, which typically refer to 
> categories such as "elongated object", "small compact object", and so 
> forth.But these straddle the boundary between grammar and lexicon, 
> and, more importantly, are typically organized paradigmatically, 
> rather than hierarchically, as is the case for animacy categories.
> -- 
David Gil

Department of Linguistic and Cultural Evolution
Max Planck Institute for the Science of Human History
Kahlaische Strasse 10, 07745 Jena, Germany

Email: gil at shh.mpg.de
Office Phone (Germany): +49-3641686834
Mobile Phone (Indonesia): +62-81281162816

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lingtyp/attachments/20181205/33040046/attachment.htm>

More information about the Lingtyp mailing list