Transparent free relatives 313
An additional fact that is potentially pertinent in the present context is that van
Riemsdijk (2012, Section 3.3), in discussing data like (27)–(29), which were originally
brought up in Grosu (2007) for a different reason (to which I return below), fails to
contest the acceptability of (28b).
For completeness, I note that Josef Bayer kindly informed me he has no objection
to this example, although he finds data that satisfy his preference for unambiguous
Case markers, such as (30), even better.^7 In any event, what matters in the present
context is that the TFR in (28b), as analyzed by van Riemsdijk, violates the Obliqueness
Hierarchy, just like the FR in (31), and the strong contrast in acceptability between the
two examples constitutes a serious problem for the Grafting-analysis.
(27) a. Mit was hat er noch nicht gerechnet?
with what has he yet not counted
‘What hasn’t he calculated with?’
b. *Wa s hat er widersprochen?
what has he contradicted
‘What has he contradicted?’
(28) a. Free Relative
Er hat mit [CPwas du gesagt hast] nicht gerechnet]DAT.
he has with what you said have not calculated
‘He did not reckon with what you said.
b. Transparent Free Relative
Sie spricht mit [CP was ich einen Idioten
she speaks with what.acc I an.acc idiot.acc
nennen würde]DAT.
call would
‘She speaks with what I would call an idiot.’
(29) a. Free Relative
*Er hat [CP was du gesagt hast] DAT nie widersprochen].
he has what.acc you said have never contradicted
‘He has never contradicted what you said.’
- Reviewer 1, who, recall, reported that s/he does not get the Bayer–van Riemsdijk-effect
in (24a–b) and (25a–b), also reports that s/he finds (28a–b) slightly degraded, although not
nearly as bad as (24c), (29a–b), or (31), and that two other native speakers corroborated this
judgment.
I have at the moment no explanation for this effect, and can only hope that the large-
scale experimental project I referred to will, if carried out, shed some light on this matter as
well. For present purposes, I confine myself to noting that this effect, just like the Bayer–van
Riemsdijk-effect, does not distinguish between FRs and TFRs, and thus poses no threat for the
analysis of TFRs I have argued for in the text.