Corpus Linguistics Case study B Levin, G Song & BTS Atkins (1997) Making sense of corpus data: a case study of verbs of sound. Int. J. Corpus Ling. 2:23-64.

Corpus Linguistics Case study

B Levin, G Song & BTS Atkins (1997) Making sense of corpus data: a case study of verbs of sound. Int. J. Corpus Ling. 2:23-64

2

Background

• Lexical semantics– word meaning related to word behaviour– Patterns of word behaviour can isolate grammatically-

relevant elements of meaning• Corpus evidence

– Provides evidence of what is possible (though not the opposite)

– Including frequency information to indicate how widespread

– May alert researcher to previously ignored or discounted properties

– Intuitions can be developed and systematically explored, misconceptions corrected

3

Case study: verbs of sound

• ie verbs which denote the emission of sound• clatter, crackle, creak, groan, gurgle, honk, hoot,

jingle, rattle, rumble, squawk, squeak, squeal, thud, whir, whistle (16)

• Chosen from ~120 verbs so labelled in Levin (1993), as a representative (heterogeneous) set, and deliberately excluding speech act verbs like speak or shout

4

Three questions

• What is it about the meaning of a verb of sound that allows us to know how it will behave?

• Why are there some properties shared by all of these verbs, and some shared by only some of them?

• How can this be the case if syntactic behaviour is semantically determined?

5

1. Intuitions and assumptions

• Based on various dictionaries, and two linguistic studies• They constitute a coherent verb class• All describe an event in which s.o./sthg. emits a sound• Each verb denotes a specific sound, implying specific

acoustic properties, eg– volume (squawk ~ whir)– pitch (squeak ~ rumble)– resonance (rattle ~ thud)– duration (gurgle ~ honk)

• Or a commonly heard sound, eg animal noise• Syntactically: intransitive in core sense, subject being the

emitter of the noise

6

2.1 Traditional dictionaries

• Typically give the sound emission as the first, basic, sense

• Often with a brief description, eg (LDOCE)– crackle ‘to make a repeated short sharp sound’– groan ‘to make a long deep sound’– whistle ‘to make a high or musical sound’

• Description of similar words are often differential, eg (LDOCE)– squeak ‘to make a very short high noise or cry that is

not loud’– squeal ‘to make a long loud high sound or cry’

7

2.1 Traditional dictionaries

• Definitions may specify typical subject or condition– clatter: heavy hard objects– crackle: … like something burning in a fire

• Or how they are made– By hitting or falling on a hard surface (thud, crack)– By two hard surfaces rubbing roughly against each

other (creak, squeak)– By repeated hitting of something (rap, rattle)

8

2.1 Traditional dictionaries

• Besides basic sound emission meaning, three other uses are often cited:– Expressing motion accompanied by sound (eg the

train rumbled across the bridge)– Introducing direct speech– Expressing the causation of the sound (eg he

crackled the newspaper, he clattered the bricks) … sound emitter is object of sound

• One transitive use widely found but rarely mentioned in dictionaries:– The fans squeaked their excitement– He whistled his disapproval

9

2.2 Linguistic studies

• Many verbs can be both transitive and intransitive: which usage is basic?

• Transitive verbs can be ergative (where patient becomes subject eg open) or accusative (where subject remains the agent eg eat)

• Basically intransitive verbs are classified as “unaccusative” or “unergative”, based on both– Syntax: what transitive uses are possible– Semantics: underlying role of subject (when intransitive)

• Unaccusative: subject is patient (eg die, fall)• Unergative: subject is agent (ie unlike ergative use of

basically transitive verb) (eg run, dance)

10

2.2 Linguistic studies

• Verbs of sound usually classified as basically intransitive – ie basic meaning is ‘to emit a sound’, not ‘to cause to

emit a sound’

• But further classification is controversial– Seems to depend on whether subject is typically

animate or not– Nevertheless, transitive use should be possible

• So what does corpus data say?

11

3. Corpus study

• 16 verbs (see above)

• Based (mainly) on examples found in the BNC

• Focus on transitive uses– Intransitive use conforms to prediction of

subject as emitter of sound– Transitive uses turn out to be more varied

than predicted

12

3.1 Sound emitter as direct object

• Causative sound examples– The nurses were clattering the tea cups– Supporters honked horns– The breeze rattled the leaves in the trees– He strode along thudding his staff into the ground at

every stride

• Causative motion examples– She clattered ice cubes into the glass– The wind creaked the door to and fro– Slowly they rumbled the big wheel across the sidewalk– She claimed the match, whistling her shot past Novotna

13

3.2 Sound emitter as subject

• Reaction objects – The girls groaned their envy– The fans squeaked their excitement– The kittens squealed their protest– The owl whistled new delight

• Message objects– Grandson Richard rumbled a reply– The pheasant flew off, squawking its alarm call– Hot on his heels galloped the herd, squealing its

demands

14

3.2 Sound emitter as subject

• Cognate objects– I groaned deep loud primal groans– The audience hooted its last hoot– The fans squealed their Beatlemania squeals at his

every move– Meg drove around whistling her shrill through-the-

teeth whistle again and again

• Direst speech complements– Looking around, she groans “I’ve had enough”– “Shakespeare!” hooted Carrie– “What are you doing?” she squawked

15

3.3 More transitive constructions

• Resultatives– The amps were so loud they would honk your hair

straight back from your skull– We forced our way through the crowds, hooting them

out of the way– One of the kettles must have whistled itself dry– It thuds staples through wood, walls, anything

• The way construction– Miss F clattered her way to the classroom– The brook gurgled its way to a refurbished mill– Tramcars rattled, roared and clanged their way along

Norfolk Street

16

3.4 Attested transitive uses of verbs of sound

Are there any patterns, and can we explain them?

17

4.1 When sound emitter is object

• Two interacting facets of a sound event might explain the complement-taking properties– Internal vs external causation

• Is the event initiated by an outside agent or does it originate in its participant?

– Internal vs external production• Is the nature of the sound an inherent feature of

the emitter, or a result of a combination of agents?

18

4.1 Internal vs external causation

• Verbs that cannot be used in causative sound construction denote internal causation– The tea kettle whistled– * The boiling water whistled the tea kettle– … caused the kettle to whistle– The rusty swings creaked– * The wind creaked the rusty swings– … caused the swings to creak

• Other verbs denote external causation– clatter, crackle, creak, honk, hoot, jingle, rattle,

rumble, squeak, squeal, thud

19

4.1 Internal vs external production

• Whistle, groan, rumble produced by a source within the sound emitter

• Jingle, rattle produced through surface contact between objects with certain properties

• Some sounds can be either, eg squeal can be from vocal tract, or caused by friction

• This distinction is relevant to determining which transitive constructions are found

20

4.1 Causative sound use

• Internally caused sounds cannot occur in situations which are externally caused– * The roller coaster squealed the children– * The driver squealed his passengers with his

reckless driving.– The driver squealed the tyres against the kerb

• Try it with the other verbs that don’t “allow” CS– groan, gurgle, squawk, whir, whistle

21

4.1 Causative motion use• OK: clatter, creak, jingle, rattle, rumble, squeak,

squeal, thud, whistle• But not: crackle, groan, gurgle, honk, hoot,

squawk, whir• Direct object must be capable of making the

noise, and a location must be specified• Verb must (again) denote externally produced

noise– He whistled the ball into the net– * He squawked the parrot into the cage

22

4.2 When sound emitter is subject

• Trivial case (cognate object), when object reiterates the nature of the sound, often permitting further specification– I groaned a deep, loud, primal groan

• Cognate objects found with 7 of our verbs:– groan, honk, hoot, squawk, squeak, squeal– But not clatter, crackle, creak, gurgle, jingle, rattle, rumble, thud,

whir, whistle

• Why (not) ?• Could you have a non-cognate object (a near synonym)?

– ? She groaned a spine-tingling moan

• Subject must be animate (construed as agentive), with control over the nature of the sound

23

4.2 When sound emitter is subject

• Other types of object• Emotion or attitude

– They groaned their envy, squealed their delight– Found with groan, honk, hoot, rumble, squeak,

squeal, whistle • Type of message

– The pheasant squawked its alarm call– Found with crackle, gurgle, honk, hoot, rattle, rumble,

squawk, squeal, whistle • Are there any semantic constraints to explain

what is/isn’t possible?

24

4.2 When sound emitter is subject

• Emotion or attitude• Type of message• Direct speech complement• In all three cases

– Subject must be animate (capable of expressing an emotion or attitude, sending a message, speaking)

– Sound must be associated with that attitude, message, content

• They hooted their derision / * indifference• “Shakespeare!” hooted Carrie – verb indicates emotion

– Sound must be internally caused• Does this analysis meet with your intuitions?

25

4.3 Other transitive constructions

• Resultative constructions– The kettled whistled itself dry

• Way constructions– The tram rattled its way down Market Street

• Subtypes of the same construction• Resultatives typically available for unergative

verbs– Way construction found for all 16 verbs, resultatives

easy to construct where they are missing (???)

26

Conclusions and criticisms

• Corpus data was used to get examples • Some patterns, groupings were observed• Data showed that there might be complexities not

previously covered by lexicographers/linguists• Assuming the data to be typical …• Explanations were well motivated

– Based on intuitions and common sense– In accord with (non-corpus based) linguistic assumptions

• Handling of negative/missing data possibly a bit iffy• An example of “traditional” linguistics backed up by

corpora