Temos gast(ad)o muito tempo nesse projeto: Variable Past Participles in Portuguese

TEMOS GAST(AD)O MUITO TEMPO NESSE PROJETO: VARIABLE PAST PARTICIPLES IN PORTUGUESE Os Participantes (PORT 5611)

Os Participantes 2

Prof. Scott Schwenter ¨  Graduate students: Eleni Christodulelis, Ashlee Dauphinais Civitello, Mark Hoff, Chelsea Pflum ¨  Undergraduate students: Tess Anderson, Megan Dollenmeyer, Anna Rose Flood, Linnea Fox, Mario Gutiérrez, Ángel Guzmán, Tavia Thompson, Alex Winton, Yarden Zer Zion ¨ 

Standard/Normative View 3

Portuguese “verbos abundantes” have two different participle forms (“particípios duplos”), one regular (–do), one irregular (often, but not always, equivalent to the 1sg present indicative) Verb

Regular Participle

Irregular Participle

pagar

pagado

pago

matar

matado

morto

gastar

gastado

gasto

ganhar

ganhado

ganho

aceitar

aceitado

aceito (BP), aceite (EP)

Standard/Normative View 4

“The regular form ... is used with ter (or haver) to form the compound perfect tenses. The irregular one is then used to form the passive voice and as an adjective” (Thomas 1969:223) “Some verbs in Portuguese have two participles, one irregular and one regular. The irregular forms, in these cases, are used as adjectives with the verb estar, while the regular forms combine with the verb ter in the perfect tenses” (Ganho and McGovern 2004:153)

Standard/Normative View 5

¨ 

There are hints at a more complicated picture, however: ¤  “Note:

We observe in contemporary Brazilian Portuguese a preference for irregular forms, especially gasto, ganho, and pago, even in construction with the verbs ter or haver” (Simões 2008:356) ¤  Implicit importance of 1sg present indicative form coinciding with irregular participle ¨ 

This preference for irregulars is even stronger in European Portuguese, as our data reveal

Prior Research 6

¨ 

Teixeira da Silva (2008): 670+ tokens from Brazil (VARSUL corpus) ¤  BUT

only TWO verbs showed variation between regular and irregular participle with ter/haver ¤  Only EIGHT total tokens of these two verbs ¨ 

Miara (2013): 843 tokens from a Brazilian newspaper ¤  BUT

only SEVEN verbs showed variation between regular and irregular participles ¤  Only 566 tokens from those seven verbs (83.2% irregular)

Prior Research 7

¨ 

¨ 

Other studies (Nevins & Rodrigues 2014) are experimental in nature: how subjects form participles from nonce verbs Mendes (2005, 2010) analyzed the overlap between perfects vs. progressives from a variationist perspective, but to our knowledge there are no prior studies of the participle alternation per se

Research Questions 8

¨ 

What is the distribution of irregular/regular participles in ter/haver + past participle constructions? ¤  Is

the normative rule stating that regular participles should occur in perfect constructions accurate? ¤  What factors condition the variation between the two participle variants? ¨ 

¨ 

Are there dialect differences between Brazil and Portugal? Has there been a diachronic shift toward more irregular participle use?

Method 9

Extracted tokens from online Corpus do Português (Davies & Ferreira) ¨  Used multiple lists of “verbos abundantes” from grammars and websites à substantial disagreement among authors! ¨  Found every verb that showed variation between regular/irregular participles with ter/haver in 20th century data (Portugal & Brazil) th century for ¨  Searched for same verbs in 19 diachronic comparison (Portugal only) ¨ 

Results 10

¨ 

N = 1077 tokens from variable verbs ¤  Irregular

participle forms = 584 (54%) ¤  Regular participle forms = 493 (46%) ¨ 

Result is contrary to most grammars: Irregular participles are MORE frequent with ter/haver than their regular counterparts

Dialectal Comparison-20th Century 11

Irregular Regular TOTALS

Brazil 164 (53%) 145 (47%) 309

Portugal 274 (68%) 128 (32%) 402

χ2 = 16.8, df = 1, p < .001

Clear tendency for greater irregular past participle use (15% higher) in Portugal than Brazil in 20th century data

Multivariate Analysis 12

What combination of predictors (= independent variables) accounts best for the variability? Predictors we coded:

Linguistic Predictors Infinitive (lexical verb) Type of Irreg Part. Auxiliary (ter/haver) TAM Person Number Polarity

Extralinguistic Predictors Dialect (Portugal vs. Brazil) Century (19/20; Portugal) Genre (Written vs. Oral)

Multivariate Analysis 13

What combination of predictors (= independent variables) accounts best for the variability? Predictors we coded:

Linguistic Predictors Infinitive (lexical verb) Type of Irreg Participle Auxiliary (ter/haver) TAM Person Number Polarity

Extralinguistic Predictors Dialect (Portugal vs. Brazil) Century (19/20; Portugal) Genre (Written vs. Oral)

Random Forest – 20th Century Data 14

Infinitive Irreg.1sg Dialect TAM Genre Auxiliary Polarity Person Number

0.00

0.05

0.10

0.15

0.20

0.25

Multivariate Analysis 15

Lexical verb corresponding to participle was by far the most significant predictor ¨  21 verbs (out of 58 included in our corpus search) showed participle variation in 20th century data. Verbs in boldface showed a significant probability for irregular participles, while underlined verbs showed a significant probability for regular participles: ¨ 

aceitar

eleger

envolver

extinguir

gastar

matar

salvar

acender

empregar

expressar

fartar

juntar

pagar

soltar

assentar

entregar

expulsar

ganhar

limpar

pegar

suspender

Multivariate Analysis 16

¨ 

¨ 

¨ 

What happens when we factor out the participle variability accounted for by individual verbs? Are there regularities across the dataset in the selection of participle type by other factors? Mixed-effects logistic regression model in R, with the predictor Infinitive corresponding to the participle as a random effect

17

Conditional Inference Tree 20th Century Data 1 Irreg.1sg p < 0.001

yes

no

2 TAM p < 0.001

5 Dialect p < 0.001

Portugal

Brazil

6 TAM p < 0.001

{future, infinitive}{imperfect, present}

{infinitive, present}{imperfect, future} 1 0.8

Node 8 (n = 80)

1 0.8

Node 9 (n = 102) irregular

0.8

Node 7 (n = 81)

irregular

1

irregular

0.8

Node 4 (n = 295) irregular

1

1 0.8

0.6

0.6

0.4

0.4

0.4

0.4

0.4

0.2

0.2

0.2

0.2

0.2

0

0

0

0

regular

0.6

regular

0.6

regular

0.6

regular

regular

irregular

Node 3 (n = 153)

0

Irregular Participle Form 18

¨ 

Most irregular participles are identical in form to the 1st singular present indicative ¤  Ela

tinha pago (‘She had paid’) ~ Eu pago (‘I pay’) ¤  Nós temos gasto (‘We have spent’) ~ Eu gasto (‘I spend’) ¨ 

Innovative participles in colloquial spoken BP are typically formed from the 1sg present indicative (Souza 2007) trazido à Tenho trago (‘I have brought’) ¤  Tenho pedido à Tenho peço (‘I have requested’) ¤  Tenho

Irregular Participle Form 19

¨ 

Our data show a clear preference for irregular participles when these are identical to the 1sg present indicative (all data)

Irregular Regular TOTALS

Irreg = 1sg pres ind 444 (69%) 198 (31%) 642

χ2 = 143, df = 1, p < .001

Irreg ≠ 1sg pres ind 140 (32%) 295 (68%) 435

Diachronic Results 20

¨ 

¨ 

Have the factors conditioning the choice of past participle changed over time? Data available in the Corpus do Português for comparison from Portugal, 19th vs. 20th centuries ¤  Insufficient

data from previous centuries for further diachronic analysis

Diachronic Results 21

Irregular Regular

19th Century 146 (42.4%) 198 (57.6%)

20th Century 273 (68.3%) 127 (31.7%)

χ2 = 50.1, df = 1, p < .001

•  Clear shift toward greater use (25.9% higher) of irregular participles between the 19th and the 20th centuries in Portugal •  Part of this shift may result from normative view that values irregular participles over their regular counterparts (Patrícia Amaral, p.c.)

Diachronic Results 22

100 19th % irregular

90

20th % irregular

80 70 60 50 40 30 20 10 0 aceitar

salvar

entregar

matar

ganhar

gastar

pagar

23

Random Forest Diachronic Data: Portugal Infinitive Irreg.1sg Century TAM Person Auxiliary Polarity Genre Number

0.00

0.05

0.10

0.15

0.20

0.25

Diachronic Results 24

Beyond the increase in overall frequency, multivariate analysis of the 19th vs. 20th century Portugal data shows another clear distinction th ¨  In the 19 century, the variation is mainly lexical in nature, but still varies by the form of irregular participle th ¨  TAM as a predictor is significant only in the 20 century ¤  Mainly

pertains to infinitive vs. imperfective, which jointly make up ≈70% of the data overall

25

Conditional Inference Tree Diachronic Data: Portugal 1 Irreg.1sg p < 0.001

no

yes

2 Century p < 0.001

20

7 Century p < 0.001

19

3 TAM p < 0.001

19

20

{infinitive, present}{imperfect, future} 1 0.8

Node 8 (n = 194)

1 0.8

Node 9 (n = 241) irregular

0.8

Node 6 (n = 172)

irregular

1

irregular

0.8

Node 5 (n = 80) irregular

1

1 0.8

0.6

0.6

0.4

0.4

0.4

0.4

0.4

0.2

0.2

0.2

0.2

0.2

0

0

0

0

regular

0.6

regular

0.6

regular

0.6

regular

regular

irregular

Node 4 (n = 81)

0

Conclusions 26

Participle choice with “verbos abundantes” does not resemble normative grammar accounts in the least! ¨  Overall, irregular participles are MORE frequent with ter/haver than regular participles ¨ 

¨ 

The most important predictor of participle choice, by far, is a lexical one, i.e. individual verb tendencies But in addition there are clear patterns of conditioning beyond the idiosyncratic effects of the individual verbs

Conclusions 27

¨ 

¨ 

¨ 

The internal grammar of participle choice in perfect constructions depends on the form of the irregular participle and on TAM Clear dialectal differentiation in 20th century: EP significantly more likely to employ irregular participles than BP Diachronic change evident in the EP data ¤  Significant

increase in the frequency of irregulars ¤  Development of TAM restriction in 20th century

Obrigado! 28

¨ 

Many thanks to Patrícia Amaral, Sonia Barnes, and especially Hannah Washington for bibliographical pointers, statistical guidance, and analytical acumen!