Au­thors: Anne Abeillé, Bar­bara Hem­forth and Elodie Winck­el

French rel­a­tive claus­es in­tro­duced by dont (from now on: dont-​RC) have been well in­ves­ti­gat­ed on the the­o­ret­i­cal level (Go­dard 1988) but have not lead to many em­pir­i­cal stud­ies (Gapa­ny 2004). Blanche-​Ben­veniste (1995) ob­served that dont-​RC are not very com­mon in spo­ken French and re­strict­ed to the ex­trac­tion of a few verb com­ple­ments. The aim of this ar­ti­cle is to de­scribe dont-​RC on the basis of three em­pir­i­cal stud­ies: two cor­pus stud­ies and one on­line ex­per­i­men­tal study. The cor­pus stud­ies have been con­duct­ed on a writ­ten cor­pus (French Tree­bank) and on a spo­ken cor­pus (Cor­pus de Français Parlé Parisien des années 2000). The re­sults con­firm part­ly Blanche-​Ben­veniste’s claim: the writ­ten usage of dont-​RC dif­fers from its spo­ken use, but it is not re­strict­ed in the later to spe­cif­ic id­ioms. Both cor­po­ra cor­rob­o­rate the work of Go­dard (1980) : dont is wide­ly used as a com­ple­ment on noun, ex­trac­tions out of NP are not marginal, not even out of sub­ject NP which are con­sid­ered non-​gram­mat­i­cal by some lin­guis­tic the­o­ries. The cor­pus stud­ies show that this kind of ex­trac­tion is wide­ly-​used in French and even con­cern “real sub­jects” of tran­si­tive verbs, which goes against the pre­dic­tion made by many gen­er­a­tive the­o­ries. In gen­er­a­tive gram­mar, rel­a­tive claus­es (as well as any ex­trac­tion) are sub­ject to is­land con­straints (Ross 1957). The nom­i­nal is­land con­straint pre­dicts that the ex­trac­tion of a noun com­ple­ment should be hard­er than the ex­trac­tion of a verb com­ple­ment. The sub­ject is­land con­straint rules out ex­trac­tions out of sub­ject NPs, which is claimed to be only ac­cept­able when the sub­ject NP is an un­der­ly­ing ob­ject (Chom­sky 2008). How­ev­er, in our cor­po­ra, dont-​RC ex­trac­tions out of NP sub­ject are more fre­quent than ex­trac­tions out of NP ob­ject. This ob­ser­va­tion con­firms Gib­son’s De­pen­den­cy Lo­cal­i­ty The­o­ry (DLT), which pre­dicts that the ex­trac­tion from sub­ject NP should have less cog­ni­tive costs than the ex­trac­tion from ob­ject NP. Other ob­ser­va­tions cor­rob­o­rate this the­o­ry as well: for ex­am­ple, that dont-​RC cor­re­spond­ing to the ex­trac­tion of the com­ple­ment of a verb favor pronom­i­nal or in­vert­ed sub­jects. Our ex­per­i­men­tal study is an on­line ac­cept­abil­i­ty task. Its aim is to com­pare the judg­ments from na­tive speak­ers re­lat­ed to dont-​RC im­ply­ing an ex­trac­tion out of sub­ject NP and ob­ject NP (the kind of sub­ject in­volved in the later case has an im­pact of the DLT’s pre­dic­tions and have been test­ed as well). The re­sults con­firm again the pre­dic­tions from the DLT: the judg­ments for the ex­trac­tion out of sub­ject NPs are sig­nif­i­cant­ly high­er than for the ex­trac­tion out of ob­ject NPs.

The full article is available here: http://www.shs-conferences.org/articles/shsconf/pdf/2016/05/shsconf_cmlf2016_14011.pdf

Data from the corpus studies and experiment are available here: https://osf.io/ebrfm/

Elodie Winckel

View all posts