How much overt agreement is needed for polysynthesis

How much overt agreement

is needed for polysynthesis?

Quantitative evidence from

Cherokee.

Erin Humphreys & Brian Hsu (UNC-Chapel Hill)

Penn Linguistics Conference 48

March 2024

Slides online: hsub.web.unc.edu

1. Overview

2. Polysynthesis & agreement

3. Cherokee word order & agreement

4. Quantitative analysis

5. Implications & conclusion

Slides online: hsub.web.unc.edu

1. Overview

Two properties that cluster in polysynthetic

languages (Jelinek 1984, Baker 1996, Mithun 2017 a.o.):

§ Flexible clausal word order

• Often with free omission of argument NPs, non-

configurational properties (Hale 1983).

§ Polypersonal agreement: Coindexation of

multiple thematic arguments in verbal agreement

morphology

Slides online: hsub.web.unc.edu

1. Overview

This paper re-examines a standard generative

proposal on the causal relationship between

these properties (Jelinek 1984; Baker 1996; 2006).

We focus on key grammatical properties of

Cherokee (Southern Iroquoian: Oklahoma and North

Carolina).

Slides online: hsub.web.unc.edu

1. Overview

Key grammatical properties of Cherokee:

(i) Highly flexible clausal word order, with known

predictive factors (Scancarelli 1987; Hsu & Frey 2023)

Example: Agent arguments of transitive predicates

can precede or follow verbs.

(1) Subject < Verb

[gitli] ogi-sdawadvs-v.

dog 1.PL.EXCL-follow-EXPP

‘The dog followed us.’ (Feeling et al. 2017; 101)

1. Overview

Key grammatical properties of Cherokee:

(i) Highly flexible clausal word order, with known

predictive factors (Scancarelli 1987; Hsu & Frey 2023)

Example: Agent arguments of transitive predicates

can precede or follow verbs.

(2) Verb < Subject

a-n-adasdelis-g-o [yvwi j-u-n-asdi].

3-PL-help-PROG-HAB people DST-3-PL-little

‘The little people help (others).’ (Feeling et al. 2017; 43)

1. Overview

Key grammatical properties of Cherokee:

(ii) Fewer NP arguments coindexed by verbal

agreement than in other polysynthetic languages

(3) [sagwu=no] [j-un-atana ahwi] d-a-hih-e.

one=CN DST-3.PL-big deer DST-3.SG-kill-REPP

‘One (of the hunters) killed big deer.’ (Feeling et al. 2017; 53)

• Features of only one 3

-person argument expressed

on the pronominal prefix a-.

1. Overview

We present a quantitative investigation of the

relation between overt coindexation of arguments

and NP placement in Cherokee clauses.

Main result: Argument NP placement does not vary,

based on whether its features are overtly coindexed.

Having fewer coindexed arguments does not lead to

reduced word order flexibility relative to other

polysynthetic languages (ex. Mohawk)

1. Overview

Proposal: The threshold for maintaining this type of word

order flexibility is abstract polypersonal agreement, rather

than surface polypersonal agreement (our terms)

Surface polypersonal agreement

Agreement morphology directly expresses features of

multiple arguments of a verb.

Abstract polypersonal agreement

Agreement morphology systematically depends

on relative properties of multiple arguments of a

verb (not all argument features directly expressed).

1. Overview

2. Polysynthesis & agreement

3. Cherokee word order & agreement

4. Quantitative analysis

5. Implications & conclusion

2. Polysynthesis & agreement

Next slides: Summary of Baker's (1996) analysis of

Mohawk (Northern Iroquoian), and the relation

between polypersonal agreement and flexible word

order.

2. Polysynthesis & agreement

Mohawk: Flexible clausal word order.

S V O

(4)a. Sak ra-núhwe’-s ako-[a]tyá'tawi

Sak MsS-like-HAB FSP-dress

'Sak likes her dress'

(Baker 1996; 10)

2. Polysynthesis & agreement

Mohawk: Flexible clausal word order.

V S O

(4)b. ra-núhwe’-s Sak ako-[a]tyá'tawi

MsS-like-HAB Sak FsP-dress

'Sak likes her dress'

(Baker 1996; 10)

2. Polysynthesis & agreement

Mohawk: Flexible clausal word order.

S O V

(4)c. Sak ako-[a]tyá'tawi ra-núhwe'-s

Sak FsP-dress MsS-like-HAB

'Sak likes her dress'

(Baker 1996; 10)

2. Polysynthesis & agreement

Mohawk: Flexible clausal word order.

V O S

(4)d. ra-núhwe’-s ako-[a]tyá'tawi Sak

MsS-like-HAB FsP-dress Sak

'Sak likes her dress'

(Baker 1996; 10)

2. Polysynthesis & agreement

Mohawk: Flexible clausal word order.

O V S

(4)e. ako-[a]tyá'tawi ra-núhwe’-s Sak

FsP-dress MsS-like-HAB Sak

'Sak likes her dress'

(Baker 1996; 10)

2. Polysynthesis & agreement

Mohawk: Flexible clausal word order.

O S V

(4)f. ako-[a]tyá'tawi Sak ra-núhwe'-s

FsP-dress Sak MsS-like-HAB

'Sak likes her dress'

(Baker 1996; 10)

2. Polysynthesis & agreement

Mohawk: Flexible clausal word order.

Other non-configurational properties (Baker 1996):

§ Absence of weak cross-over effects

§ Free omission of non-pronominal NP arguments

2. Polysynthesis & agreement

Mohawk: Verbal agreement preﬁxes coindex

features of up to two thematic arguments (including

two 3

-person arguments)

§ Shako- expresses gender, number properties of two

arguments (masc. sg. subject, fem. object)

(5) Shako-nuhwe’-s (ne owira’a) Mohawk (Baker 1996; 21)

MSS/FO-like-HAB NE baby

‘He likes them (babies)’

2. Polysynthesis & agreement

Mohawk: Productive noun incorporation (cf. Mithun

2017b), in complementary distribution with

pronominal object agreement

(6) Ra-wir-a-nuhwe’-s Mohawk (Baker 1996; 21)

MSS-baby-ø-like-HAB

‘He likes babies’

§ Baker (1996): Incorporated objects ﬁll the

"slot" taken by object agreement.

2. Polysynthesis & agreement

Baker's proposed condition on agreement, argument

licensing in polysynthetic languages:

The Morphological Visibility Condition (MVC)

A phrase X is visible for θ-role assignment from a head Y

only if it is coindexed with a morpheme in the word

containing Y via:

An agreement relationship, or

A movement relationship

(Baker 1996; 17)

2. Polysynthesis & agreement

Baker (1996): If this condition is met, all non-

pronominal NPs pattern as adjuncts: they can be

freely omitted, freely ordered.

[In a certain class of languages], a verb [Y] agrees with an

overt NP [X] if and only if [X] is in a dislocated, adjunct

position.

(Baker 2006: 290)

2. Polysynthesis & agreement

Verbs do not overtly coindex all argument features in

all polysynthetic languages with ﬂexible word order.

How does the MVC accommodate this?

§ The MVC tolerates some degree of phonologically

null agreement (Baker 1996):

o Some argument features, or combinations thereof,

can be expressed by a null (∅) agreement morpheme.

o i.e. Particular cells in the paradigm are expressed as ∅

2. Polysynthesis & agreement

Verbs do not overtly coindex all argument features in

all polysynthetic languages with ﬂexible word order.

How does the MVC accommodate this?

§ The MVC tolerates some degree of phonologically

null agreement (Baker 1996):

o The structural position of arguments indexed by null

(∅) agreement morphemes does not diﬀer from

arguments indexed overtly (Baker 2006).

2. Polysynthesis & agreement

Our question: How much overt agreement is

necessary to maintain general word order freedom,

general polysynthetic characteristics?

Next sections: Quantitative evidence from Cherokee

shows that abstract polypersonal agreement is

sufficient to satisfy the MVC.

1. Overview

2. Polysynthesis & agreement

3. Cherokee word order & agreement

4. Quantitative analysis

5. Implications & conclusion

3. Cherokee word order & agreement

Language background: Southern Iroquoian

language, once spoken throughout lower

Appalachian regions.

Most speakers (~2000 native/ﬂuent) now reside in

Oklahoma and Western North Carolina.

Next slides: Key aspects of clausal word order,

agreement morphology.

3. Cherokee word order & agreement

Flexible clausal word order:

§ All orders of subject, object, verb are grammatical

(Pulte & Feeling 1975, Scancarelli 1987).

(7) Subject < Verb

[gitli] ogi-sdawadvs-v.

dog 1.PL.EXCL-follow-EXPP

‘The dog followed us.’ (Feeling et al. 2017; 101)

3. Cherokee word order & agreement

Flexible clausal word order:

§ All orders of subject, object, verb are grammatical

(Pulte & Feeling 1975, Scancarelli 1987).

(8) Verb < Subject

a-n-adasdelis-g-o [yvwi j-u-n-asdi].

3-PL-help-PROG-HAB people DST-3-PL-little

‘The little people help (others).’ (Feeling et al. 2017; 43)

3. Cherokee word order & agreement

Flexible clausal word order:

§ All orders of subject, object, verb are grammatical

(Pulte & Feeling 1975, Scancarelli 1987).

(9) Object < Verb

No kil [am] ji-todis-g-o

Then until water 1-heat.water-PROG-HAB

‘Then I heat some water.’ (Feeling et al. 2017; 129)

3. Cherokee word order & agreement

Flexible clausal word order:

§ All orders of subject, object, verb are grammatical

(Pulte & Feeling 1975, Scancarelli 1987).

(10) Verb < Object

u-sdu-hnv [galohisdi?i].

3-close-EXPP door

‘(he) closed the door.’ (Feeling et al. 2017; 35)

3. Cherokee word order & agreement

Flexible clausal word order:

§ Non-configurational properties shared with

Mohawk (Beghelli 1996)

• Absence of weak cross-over effects

• Free omission of non-pronominal NP arguments

3. Cherokee word order & agreement

Recent ﬁndings (Hsu & Frey 2023):

§ No subject/object asymmetry in word order:

o NP placement insensitive to number of NPs in clause:

o Ex: Themes of transitive verbs show same placement

as themes of unaccusative verbs.

§ Placement of each NP determined probabilistically

by its properties related to contrast, referential

accessibility and thematic role.

3. Cherokee word order & agreement

Recent ﬁndings (Hsu & Frey 2023):

§ Word order principles qualitatively similar

to clausal word order in related Nothern Iroquoian

languages:

• Mohawk (Baker 1996; Mithun 2017b)

• Cayuga (Mithun 1992)

• Tuscarora (Mithun 1995)

3. Cherokee word order & agreement

Pronominal agreement system:

§ All verbs have one pronominal preﬁx

o Person and number properties of at least one

argument are indexed

o No paradigmatic gaps, all combinations of person and

number are available.

§ If both arguments are local (1st person, 2nd

person), features of both, and their relative

thematic prominence, are expressed by the preﬁx

3. Cherokee word order & agreement

Pronominal agreement system:

§ If both arguments are local (1

or 2

person): features of both, and their relative

thematic prominence, are expressed by the preﬁx

(11) hi? da-jv-nej-el-i.

this FUT-1SG/2PL-tell-DAT-FUT

‘I will tell you about this’

(Feeling et al. 2017; 75)

3. Cherokee word order & agreement

Pronominal agreement system:

§ If one argument is local, and the other is not: the

preﬁx expresses features of the local argument

o Regardless of relative thematic prominence of the

arguments

3. Cherokee word order & agreement

Pronominal agreement system:

§ If one argument is local, and the other is not: the

preﬁx expresses features of the local argument

o Regardless of relative thematic prominence of the

arguments

(12) [gitli] ogi-sdawadvs-v.

dog 1.PL.EXCL-follow-EXPP

‘The dog followed us.’

(Feeling et al. 2017; 101)

3. Cherokee word order & agreement

Pronominal agreement system:

§ If both arguments are 3rd person: The prefix

expresses number features of the more animate

argument

o If they are equally animate, the more thematically

prominent argument is overtly agreed with

(13) [sagwu=no] [j-un-atana ahwi] d-a-hih-e.

one=CN DST-3.PL-big deer DST-3.SG-kill-REPP

‘One (of the hunters) killed big deer.’

(Feeling et al. 2017; 53)

3. Cherokee word order & agreement

Pronominal agreement system:

§ If both arguments are 3rd person: The preﬁx

expresses number features of the more animate

argument.

o If they are equally animate, the more thematically

prominent argument is overtly agreed with

(14) [ijul] y-uni-kevs-e

both NONFIN-3.PL-chase-REPP

‘(The bull) would chase them both.’

(Feeling et al. 2017; 114)

3. Cherokee word order & agreement

§ Mohawk: Features of two 3rd person arguments

can be expressed on a pronominal preﬁx.

o (Baker 1996; 191; Mithun 2017b; 33)

§ Cherokee: Features of two 3rd person arguments

are never simultaneously expressed on a preﬁx.

o Pronominal preﬁxes on transitive verbs are only

partially informative about the features of the non-

coindexed argument.

o Ex. A 3rd person preﬁx on a transitive verb only

indicates that the other argument is not local

3. Cherokee word order & agreement

§ Cherokee only consistently expresses features of

one argument of a transitive verb.

o Unlike Mohawk

§ There is no productive noun incorporation

o Also unlike Mohawk

1. Cherokee word order & agreement

A proposed distinction among polypersonal

agreement systems:

Surface polypersonal agreement

Agreement morphology directly expresses features of

multiple arguments of a verb

Abstract polypersonal agreement

Agreement morphology systematically depends

on relative properties of multiple arguments of a

verb (not all argument features directly expressed)

3. Cherokee word order & agreement

Pronominal preﬁxes in Cherokee typically express

features of only one argument of transitive verb.

§ Few patterns of surface polypersonal

agreement, compared to Mohawk.

However, the choice of which argument to coindex

depends on relative properties of more than one

thematic argument.

§ Evidence for abstract polypersonal agreeement.

3. Cherokee word order & agreement

Next section: Does this reduced amount of visible

argument features on verbs result in a distinct

structural representation of NPs in the clause?

Or, is abstract polypersonal agreement suﬃcient to

maintain word order ﬂexibility?

3. Cherokee word order & agreement

Interim summary

§ Cherokee has less overt coindexation than the

other Iroquoian languages

o Taken as the "prototypical examples of polysynthesis"

(Mithun 2017b; 747)

§ We know what predicts word order in Cherokee

We use Cherokee to test hypotheses about the

relationship between word order flexibility and

morphology

3. Cherokee word order & agreement

If Surface Polypersonal Agreement is needed to satisfy

the MVC...

Whether or not an argument is overtly agreed with will

signiﬁcantly aﬀect its position in the clause.

If Abstract Polypersonal Agreement meets the

agreement threshold for the MVC...

Whether an argument is overtly agreed with will have no

signiﬁcant eﬀect on its position in the clause.

1. Overview

2. Polysynthesis & agreement

3. Cherokee word order & agreement

4. Quantitative analysis

5. Implications & conclusion

4. Quantitative analysis

Hsu and Frey (2023) identify factors that best predict

clausal word order in Cherokee:

§ All non-pronominal NPs in a corpus are tagged for a

range of grammatical properties.

Two word order factors found to be signiﬁcant in

predicting the placement of NPs:

§ Thematic role

§ Referential accessibility (a.k.a. newness / givenness)

4. Quantitative analysis

Probabilistic tendencies of Cherokee word order in

relation to thematic role and referential accessibility:

(Hsu & Frey 2023)

NP is an agent

NP is a theme

NP is new

92%

preverbal

(12/13)

73%

preverbal

(36/49)

NP is given

76%

preverbal

(37/49)

50%

preverbal

(57/114)

4. Quantitative analysis

This paper: We expand the Hsu & Frey corpus by

tagging all NPs for whether their features are

coindexed by a pronominal prefix on the verb.

We can thus identify whether overt coindexation on

a pronominal prefix influences placement of NP

arguments of transitive verbs.

§ On intransitive verbs, features of the sole

argument NP are always coindexed.

4. Quantitative analysis

We control for factors that inﬂuence NP placement,

and/or argument coindexation:

§ Thematic role aﬀects word order and

coindexation

§ Argument animacy aﬀects only coindexation

§ Referential accessibility aﬀects only word order

(Montgomery-Anderson 2015; Hsu & Frey 2023)

4. Quantitative analysis

Agent NPs are coindexed significantly more than

themes are. It also has a word order effect: agent NPs

are likelier to precede verbs than themes (H&F 2023)

X-squared = 67.803,

df = 1,

p-value < 2.2e

-16

4. Quantitative analysis

When controlling for thematic role: No

signiﬁcant eﬀect of coindexation on NP position.

X-squared = 0.040695,

df = 1,

p-value = 0.8401

Figure: Placement of theme NPs

only, by coindexed vs. non-

coindexed status.

4. Quantitative analysis

Animacy signiﬁcantly aﬀects which arguments are

coindexed but not word order.

X-squared = 170.01,

df = 2,

p-value < 2.2e-16

4. Quantitative analysis

When controlling for animacy contrasts: No

signiﬁcant eﬀect of coindexation on NP position.

X-squared = 0.021634,

df = 1,

p-value = 0.8831

Figure: Placement of animate NPs

only, by coindexed vs. non-

coindexed status.

4. Quantitative analysis

Referential accessibility does not signiﬁcantly aﬀect

which argument is indexed. It does have word order

eﬀects; discourse-new NPs likelier to precede verbs

than given NPs (H&F 2023)

X-squared = 0.5467,

df = 1,

p-value = 0.4597

4. Quantitative analysis

When controlling for referential accessibility: No

significant effect of coindexation on NP position.

X-squared = 0.0038223,

df = 1,

p-value = 0.9507

Figure: Placement of discourse-

given NPs only, by coindexed vs.

non-coindexed status.

4. Quantitative analysis

The placement of an argument NP in relation to a

verb does not depend on whether its features are

visibly coindexed by a pronominal preﬁx

In Cherokee, non-coindexed arguments appear to

have the same structural status as coindexed

arguments

1. Overview

2. Polysynthesis & agreement

3. Cherokee word order & agreement

4. Quantitative analysis

5. Implications & conclusion

Cherokee does not have less word order freedom

than other polysynthetic languages, while showing

overt agreement with systematically fewer

arguments.

Abstract polypersonal agreement is suﬃcient to meet

the agreement threshold for maintaining general

polysynthetic properties.

5. Implications & conclusion

What does this mean for the MVC?

Baker (2006): Nonvisibly-coindexed NPs show flexible

placement if their features happen to be expressed by a

null affix (∅ represents a slot in the agreement paradigm)

Cherokee agreement is hard to characterize as such:

Whether features of an NP are coindexed depends on

relative properties of two arguments.

5. Implications & conclusion

What does this mean for the MVC?

Crucially, verbal agreement in Cherokee requires all NP

arguments to be visible to the verb, in order to determine

which features are coindexed.

§ Can be modeled by Multiple Agree between verbs and

argument NPs (Oxford 2019).

A one argument to one aﬃx correspondence is not

strictly required.

5. Implications & conclusion

The Morphological Visibility Condition (MVC)

A phrase X is visible for θ-role assignment from a head

Y only if it is coindexed with a morpheme in the

word containing Y via:

An agreement relationship*, or

A movement relationship

(Baker 1996; 17)

*abstract polypersonal agreement is sufficient to meet

this threshold

5. Implications & conclusion

Summary of main results:

§ NP placement in the Cherokee clause does not depend

on whether its features are visibly coindexed on a

pronominal preﬁx.

§ Abstract polypersonal agreement is suﬃcient to

maintain free word order in polysynthetic languages

5. Implications & conclusion

Future directions:

§ More quantitative analysis of word order predictors in

polysynthetic languages and/or languages with ﬂexible

word order.

§ How much can languages vary in the movement-

triggering propensities of particular features? Is it

predicted by any (morphological) properties?

Thank you!

Slides online: hsub.web.unc.edu

References

Adger, David., Daniel Harbour, and Laurel J. Watkins. 2009. Mirrors and

macroparameters. Cambridge University Press.

Akkus, Faruk. 2018. Copular constructions and clausal syntax in Cherokee.

In Proceedings of the Workshop on the Structure and Constituency of

Languages of the Americas 21, University of British Columbia Working

Papers in Linguistics 46, ed. Megan Keough, Natalie Weber, Andrei

Anghelescu, Sihwei Chen, Erin Guntly, Khia Johnson, Daniel Reisinger,

and Oksana Tkachman.

Beghelli, Filippo. 1996. Cherokee clause structure. In Cherokee papers from

UCLA, ed. Filippo Beghelli, Barbara Blankenship, Michael Dukes, Edward

S. Flemming, Pamela Munro, Brian Potter, Robert S. Williams, and

Richard Wright, 105–114. Los Angeles: Department of Linguistics,

University of California Los Angeles.

Baker, Mark. 1996. The polysynthesis parameter. Oxford: Oxford

University Press.

References

Baker, Mark. 2006. On zero agreement and polysynthesis. In Arguments

and Agreement, ed. Peter Ackema, Patrick Brandt, Maaike

Schoorlemmer, and Fred Weerman, 289-342. Oxford Univeristy Press.

Chomsky, Noam. 1993. A minimalist program for linguistic theory. In The

View from Building 20, ed. Kenneth Hale and Samuel Jay Keyser, 1–52.

Cambridge, MA: MIT Press.

Dipper, Stefanie, Michael Götze, and Stavros Skopeteas, ed. 2007.

Information structure in cross-linguistic corpora: Annotation guidelines

for phonology, morphology, syntax, semantics, and information

structure (ISIS Working Papers of the SFB 632). Potsdam:

Universitätsverlag.

Feeling, Durbin, and William Pulte. 1975. Cherokee-English dictionary.

Talequah: Cherokee Nation of Oklahoma.

References

Frey, Benjamin. 2020. “Data is nice:” Theoretical and pedagogical

implications of an Eastern Cherokee corpus. In Collaborative

approaches to the challenge of language documentation and conservation:

Selected papers from the 2018 Symposium on American Indian Languages

(SAIL), ed. Wilson de Lima Silva and Katherine Riestenberg, 38–53.

Honolulu: University of Hawai’i Press.

Hale, Kenneth. 1983. Warlpiri and the grammar of non-configurational

languages. Natural Language & Linguistic Theory 1, 5–47.

Harbour, Daniel, Laurel J. Watkins, and David Adger. 2012. Information

structure, discourse structure, and noun phrase position in Kiowa.

International Journal of American Linguistics 78: 97–126.

Hsu, Brian, and Benjamin Frey. 2023. Word order in Cherokee:

information structure, thematic structure, and variability. Ms.,

University of North Carolina at Chapel Hill.

References

Jelinek, Eloise. 1984. Empty categories, case, and conﬁgurationality. NLLT

2: 39–76.

Jelinek, Eloise. 2006. The Pronominal Argument Parameter. In Arguments

and Agreement, ed. Peter Ackema, Patrick Brandt, Maaike

Schoorlemmer, and Fred Weerman, 261–288. Oxford University Press.

Mithun, Marianne. 1992. Is basic word order universal? In Pragmatics of

word order ﬂexibility, ed. Doris L. Payne, 15–62. Amsterdam: John

Benjamins.

Mithun, Marianne. 1995. Morphological and prosodic forces shaping word

order. In Word order in discourse, ed. Pamela A. Downing and Michael

Noonan, 387–423. Amsterdam/Philadelphia.

Mithun, Marianne. 2017a. Argument marking in the polysynthetic verb

and its implications. In The Oxford handbook of polysynthesis, ed.

Michael Fortescue, Marianne Mithun, and Nicholas Evans, 30–58.

Oxford University Press

References

Mithun, Marianne. 2017b. The Iroquoian language family. In The

Cambridge handbook of linguistic typology, ed. Alexandra Y. Aikhenvald,

747–781. Cambridge University Press.

Montgomery-Anderson, Brad. 2008. A reference grammar of Oklahoma

Cherokee. Ph.D dissertation. University of Kansas.

Montgomery-Anderson, Brad. 2015. Cherokee Reference Grammar.

Norman: University of Oklahoma Press.

Oxford, Will. 2019. Inverse marking and multiple agree in algonquin:

Complementarity and variability. Natural Language & Linguistic

Theory, 37(3), 955–996.

Posit team. 2023. RStudio: Integrated Development Environment for

R. Posit Software, PBC, Boston, MA. URL http://www.posit.co/.

References

R Core Team. 2023. R: A Language and Environment for Statistical

Computing. R Foundation for Statistical Computing, Vienna, Austria.

<https://www.R-project.org/>.

Scancarelli, Janine. 1986. Pragmatic roles in Cherokee grammar. In

Proceedings of the Twelfth Annual Meeting of the Berkeley Linguistics

Society, ed. Vassiliki Nikiforidou, Mary VanClay, Mary Niepokuj, and

Deborah Feder, 224–234. Berkeley: Berkeley Linguistics Society.

Wickham, H. et al. (2019). Welcome to the tidyverse. Journal of Open Source

Software, 4(43), 1686, doi:10.21105/joss.01686