\documentclass[letter,11pt]{article} %UVIC Style-sheet \usepackage[utf8x]{inputenc} \usepackage[T1]{fontenc} \usepackage{times} \usepackage[hmargin=1.75in,vmargin=1.75in]{geometry} \usepackage{sectsty} %package goes with \sectionfont commands below, ensures they are same font size as text. %\usepackage{indentfirst}% indents first paragraph in each section, as per style sheet %\usepackage{titlesec} \usepackage[pagestyles]{titlesec} % related to title spacing commands below. \usepackage{ragged2e}%allows word hyphenation at the end of lines for ragged-right text \usepackage{expex} \usepackage{graphics} \usepackage[pdftex]{graphicx} \usepackage{wasysym,pgfpages} \usepackage{qtree} \usepackage{amsfonts} \usepackage{multirow} %for merging cells in tabular \usepackage{tabularx} \usepackage{textcomp} \usepackage[T1]{tipa} \usepackage[normalem]{ulem} \usepackage{natbib,natbibspacing} %\usepackage{fancyhdr} \newcommand{\g}[1]{\protect\tipaUpperaccent[.1ex]{\lower.7ex\hbox{\textipa{'}}}{#1}} \newcommand{\gx}[1]{\protect\tipaUpperaccent[.1ex]{\lower.4ex\hbox{\textipa{'}}}{#1}} %\renewcommand{\large}[1]{\fontsize{11}{11}} \sectionfont{\normalsize} \subsectionfont{\normalsize} \subsubsectionfont{\normalsize} %goes with \sectsty package, suppresses default large font size on subsections % \setlength{\parskip}{\baselineskip} \definelingstyle{ubcwpl}{glstyle=wrap,glhangindent=.25in,glhangstyle=progressive,glhangrightskip=.37,interpartskip=10pt,glspace=.6em,everygla=,everyglb=,everyglc=,everygl=,everyglft=,aboveglftskip=0pt,aboveexskip=10pt,belowexskip=10pt,labeloffset=1ex,textoffset=1ex} %defines UBCWPL example style, specifically includes progressive hanging indentation and line skips. \lingset{lingstyle=ubcwpl}%used with expex package to set default example formatting %\lingset{glstyle=wrap,labeltype=alpha,interpartskip=10pt,glspace=.6em,glhangindent=.25in,glhangstyle=normal,everygla=,everyglb=,everyglc=,everygl=,everyglft=,aboveglftskip=0pt,aboveexskip=10pt,belowexskip=10pt} %used with expex package to set default example formatting \newlength{\aftersection} \setlength{\aftersection}{11pt} \newlength{\beforesection} \setlength{\beforesection}{11pt} %ensures that default spacing before and after sections is equal to 1 (10pt) line \titlespacing*{\section}{0pt}{\beforesection}{\aftersection} \titlespacing*{\subsection}{0pt}{\beforesection}{\aftersection} \titlespacing*{\subsubsection}{0pt}{\beforesection}{\aftersection} \pagestyle{empty} %gets rid of page numbers % \newpagestyle{mystyle}{% % %\headrule\sethead{\chaptername\ \thechapter. \chaptertitle}{}{} % \setfoot{}{\footnotesize{\emph{Working Papers of the Linguistics Circle of the University of Victoria} 21-2\\ % \emph{\textcopyright 2011 John Lyon}}}{}} % \pagestyle{mystyle} \begin{document} \let\eachwordone=\rm %supresses italics in gb4e examples, if that's what you're using. \begin{center}{\fontsize{15pt}{\baselineskip}{\bf{Oblique Marked Relatives in Southern Interior Salish: Historical Implications for a Movement Analysis\let\thefootnote\relax\footnotetext{My research has been supported through grants from the Jacobs Research Fund and the American Philosophical Society. I wish to thank my main Okanagan consultants, Lottie Lindley and Sarah McLeod, for their patience. Thanks also to Ewa Czaykowska-Higgins, Henry Davis, Dwight Gardiner, and Karsten Koch for helpful feedback and for providing data.}}}}\end{center} \begin{center}{\fontsize{12pt}{\baselineskip}{\textbf{John Lyon}}\\{University of British Columbia}\\ \emph{metalinguist@gmail.com}}\end{center} \vspace{10pt} {\addtolength{\leftskip}{0.5in} \addtolength{\rightskip}{0.5in} \noindent{\small{This paper investigates the distribution of the oblique marker \emph{t} in relative clauses in the Upper Nicola dialect of Okanagan and Nxa\textglotstop amxcín \citep{Willett:2003}, specifically in light of relative clauses in the Northern Interior Salish languages St'\'at'imcets \citep{Davis:2004,Davis:2010a} and {N\textbeltl\textschwa\textglotstop kepmxc\'in} \citep{Kroeber:1997,Kroeber:1999,Koch:2004,Koch:2006} which show evidence for movement of a clause-internal DP to the left periphery of CP. Data from Southern Interior Salish languages also show evidence for clause-internal movement, but the distribution of the oblique marker suggests that the landing site of the moved DP is in a higher position. This distinction between Northern and Southern Interior Salish may be construed as evidence for a historical split with regards to relative clause formation, and may have occurred at roughly the same time as the inversion of prepositions to a DP-internal position in the Southern Interior.}} \par} \vspace{10pt} %\RaggedRight goes with ragged2e package \setlength{\parindent}{0.38in} \section{\hspace*{.16in}Introduction} Okanagan, Nxa\textglotstop amxcín (a.k.a. Moses-Columbian), Coeur d'Alene and the dialect continuum known as Spokane-Kalispel-Flathead comprise the Southern Interior sub-branch of the Salish language family. Okanagan is spoken in South-central British Columbia and North-central Washington. It is critically endangered, being spoken by only about 400 speakers. The Upper Nicola dialect of Okanagan is centered around the Douglas Lake (Sp\'a\v{x}m\textschwa n) and Quilchena (N\textbeltl\g{q}\'i\textbeltl m\textschwa lx) reserves, close to the city of Merritt, B.C., by perhaps as few as 12 speakers. Nxa\textglotstop amxcín is spoken in central Washington, primarily in Colville territory, by fewer than forty speakers \citep[3]{Willett:2003}. Southern Interior Salish languages have syntactic structures which may be described as relative clauses, in the sense that these clauses contribute information which further specifies the referent of a head noun \citep{Kroeber:1999}. A case may also be made for a more formal definition, where a relative clause consists of a ``syntactically complex modifier involving abstraction over an internal position of the clause (the relativization site) and connected to some constituent it modifies (the relative ``head'')'' \citep{Bianchi:2002}. Example (\ref{rca}) shows a typical Okanagan relative in brackets. The determiner \emph{i\textglotstop} and the oblique marker \emph{t} introduce the clausal remnant \emph{{\v{x}\textsuperscript{w}\'ilst\textschwa m} {$t_1$}} ``he/she was abandoned by x''. This clause modifies an NP head \emph{sqilx\textsuperscript{w}} ``people''.\footnote{Similar to other branches of the family, Southern Interior Salish languages lack a dedicated relative pronoun or complementizer.}$^,$\footnote{The determiner-oblique marker sequence \emph{i\textglotstop} \emph{t}, in combination with the ``passive'' suffix \emph{-m} indicate that it is the agent of the passive sentence (i.e. \emph{sqilx\textsuperscript{w}} ``people'') which has been extracted. Since main-clause passive agent nominals are introduced by \emph{i\textglotstop} \emph{t} and occur post-predicatively, the DP \emph{i\textglotstop} \emph{t} \emph{sqilx\textsuperscript{w}} in (\ref{rca}) must have raised from a post-predicative position at some point during the derivation. See Koch (2006) and Davis (2010) for a discussion of evidence pertaining to whether a matching or raising analysis is correct for N\textbeltl\textschwa\textglotstop kepmxc\'in and St'\'at'imcets, respectively. Davis, for instance, concludes that some types of relatives in St'\'at'imcets require a matching analysis \citep{HulseySauerland:2006}, however there is no evidence for raising in any St'\'at'imcets relatives. For the purposes of this paper, I assume that Okanagan patterns similarly to St'\'at'imcets, and therefore adopt the matching analysis as the null hypothesis. The subscript `2' indicates that the external head noun is co-referent with the RC-internal one, which undergoes deletion through identity.} \ex[textoffset=.18in]\label{rca} \begingl \gla {ix\'i\texthalflength\textglotstop,} {u\textbeltl } \textbf{{i\textglotstop }} \textbf{sqilx\textsuperscript{w}$_2$} [[\textbf{i\textglotstop } \textbf{t} \textbf{[${\oslash}_{NP_2}$]$_{DP_1}$]} \textbf{\v{x}\textsuperscript{w}\'ilst\textschwa m} {\emph{\textbf{$t_1$}$_{CP}$}]} {k\textsuperscript{w}uk\textsuperscript{w}} {c\'utl\textschwa x} ``{wa\g{y}} {cak\textsuperscript{w}} {\textglotstop aws\textrevglotstop \'a\g{c}nt\textschwa m} {mat} {sti\g{m}} {i\textglotstop } {c\'awts.}''// \glb \textsc{dem} \textsc{conj} \textsc{det} people \textsc{det} \textsc{obl} { } abandon-\textsc{caus-pass} {\emph{gap}} \textsc{evid} say-\textsc{3pl.abs} yes \textsc{deon} go-look-\textsc{dir-1pl.erg} \textsc{evid} what \textsc{det} doings-\textsc{3sg.poss}// \glft Meanwhile, \textbf{the people who abandoned him}, they said ``We should go see what he's doing.'' (from Upper Nicola legend)// \endgl \xe %6042txtLL81[legend] It has been well-established that relative clauses in the Northern Interior Salish languages N\textbeltl\textschwa\textglotstop kepmxc\'in (a.k.a. Thompson) \citep{Kroeber:1997,Kroeber:1999, Koch:2006} and St'\'at'imcets (a.k.a. Lillooet) \citep{Davis:2004,Davis:2010a} are formed by movement of a clause-internal DP to the left-periphery of the relative clause CP, but besides \citet{Kroeber:1999}, and a chapter in \citet{Willett:2003} on relative clauses in Nxa\textglotstop amxcín, little work has been done on relativization strategies of the Southern Interior. As illustrated by the bracketing in (\ref{rca}), I claim that clause internal movement also occurs in the formation of Okanagan relatives.\footnote{Okanagan relatives, like St'\'at'imcets relatives \citep{Davis:2010a}, show evidence for an A' dependency within the relative clause \citep{Chomsky:1977}: resumptive pronouns are not permitted clause-internally, and long range extraction is possible, subject to strong island effects. For reasons of space, I do not include these data.} This paper investigates several points pertaining to relativization in Okanagan and the Southern Interior. First, I show that the formation of at least some relative clauses in Okanagan (and at least one other Southern Interior language, Nxa\textglotstop amxcín) involves movement of a clause-internal DP to the left-periphery of the relative clause. Secondly, I show that certain classes of oblique-marked relative clauses in Okanagan and Nxa\textglotstop amxcín, which at first seem to defy a movement analysis, are explained if the moved DP lands in a higher position than the Spec CP position argued for by \citet{Davis:2010a} for St'\'at'imcets. I claim that this difference represents a more general split between relative clause formation in the Northern and Southern Interior languages. Finally, I suggest that diachronically, there is a causal relation between the DP-internal ``prepositions'' characteristic of languages in the Southern Interior, and the structure of relative clauses in these languages. More specifically, inversion of prepositions to a DP-internal position may have conditioned a change in relative clause formation in the Southern Interior. %CHECK more general split, is there any data from Kalispel/ coeur d'alene? The paper is organized as follows: Section 2 discusses some basic facts about Okanagan DP structure, and introduces relative clauses. Section 3 summarizes the theory of relative clause formation by movement, and presents data showing that certain classes of relative clauses in Okanagan support a movement analysis for this language. Section 4 presents Okanagan data involving certain types of oblique-marked relative clauses which are problematic for the movement analysis, and then discusses similar data in other Interior Salish languages, which prove illuminating to the problem at hand. Section 5 presents my solution to this problem. Section 6 discusses further historical implications of this analysis. Section 7 raises further questions, and section 8 concludes. \section{\hspace*{.16in}Introducing Okanagan relatives} \subsection{\hspace*{.04in}Okanagan DP Structure} Okanagan, like other Salish languages, is verb-initial, however in transitive sentences involving two overt nominal arguments, subject-verb-object (SVO) is an unmarked word order. The language exhibits a tight correlation between predicate transitivity and argument marking. While subject nominals will always be introduced by a determiner \emph{i\textglotstop}, object nominals are only introduced by \emph{i\textglotstop} if the predicate is formally transitive, as in (\ref{1a}a) \citep{Lyon:2011}. If the predicate is formally intransitive, an object nominal will always be introduced by the oblique marker, as in (\ref{1a}b).\footnote{By formally transitive, I refer to predicates which are affixed by any one of several transitivizers: \emph{-nt-} `directive', \emph{-st-} `causative', \emph{-cit-} `transitive applicative', \emph{-\textbeltl t-} `ditransitive applicative'. Such predicates take ergative subject morphology. For the purposes of this paper, `formally transitive' also subsumes transitive nominalized possessive predicates, i.e. those predicates with possessor subjects, which may take a nominal DP object. By formally intransitive, I refer to predicates which are affixed by one of several intransitivizers: \emph{-\textschwa m} `middle', \emph{-(a\textglotstop)x} `intransitive'. Such predicates take absolutive subject morphology, and oblique-marked objects, never full DPs.} \pex[labeloffset=.20in,textoffset=1.5ex]\label{1a} \a\begingl \gla {tk\'ic-\textschwa n} {[\textbf{i\textglotstop} } {sq\textschwa ltm\'ix\textsuperscript{w}$_{DP}$]} {\textrevglotstop apn\'a\textglotstop } {s\v{x}\textschwa l\v{x}\textrevglotstop \'alt.}// \glb meet-\textsc{(dir)-1sg.erg} \textsc{det} man now today// \glft I met a man today.// %2249LL0[det] \endgl \a\begingl \gla {kn} {tk\'ic-\textschwa m} [\textbf{t} {sq\textschwa ltm\'ix\textsuperscript{w}$_{KP}$]} {\textrevglotstop apn\'a\textglotstop } {s\v{x}\textschwa l\v{x}\textrevglotstop \'alt.}// \glb \textsc{1sg.abs} meet-\textsc{mid} \textsc{obl} man now today// \glft I met a man today.//%2248LL1[det] \endgl \xe Intransitive objects like (\ref{1a}b) are not DPs, since the oblique marker \emph{t} is not a determiner \citep{Lyon:2011}. In specific grammatical environments, the determiner \emph{i\textglotstop} and the oblique marker \emph{t} may co-occur, as when marking an instrument (\ref{1b}a). The determiner \emph{i\textglotstop} also co-occurs with the locative markers \emph{\gx{k}l} `to/towards', \emph{l} `at/on/in', and \emph{tl} `from/than' (\ref{1b}b).\footnote{See \citet[117]{Mattina:1973} for further description of these particles.} Together, these yield a structure resembling an English prepositional phrase except that the `preposition' occurs internal to the DP \citep[71]{Kroeber:1999}.\footnote{This \textsc{det-prep} ordering is a general feature of all languages in Southern Interior Salish, and contrasts with the \textsc{prep-det} ordering exhibited by the rest of the family. \citet[72]{Kroeber:1999} hypothesizes that ``this peculiarity is readily explained if articles in these languages derive diachronically from demonstrative particles outside DP, or loosely adjoined to it, rather than from articles occupying the determiner slot within DP.'' I suggest an alternative analysis which is somewhat at odds with Kroeber's.} \pex[labeloffset=.20in,textoffset=1.5ex]\label{1b} \a\begingl \gla {\g{t}\textrevglotstop ap\textschwa nt\'is} {[\textbf{i\textglotstop} } {[\textbf{t}} {swlwlm\'ink$_{KP}$]$_{DP}$].}// \glb shoot-\textsc{dir-3sg.erg} \textsc{det} \textsc{obl} gun// \glft He shot it with a gun.// %2593LL0[instrumental] \endgl \a\begingl \gla {John} {np\'us\textschwa s} {[\textbf{i\textglotstop} } {[\textbf{l}} {\textbeltl kap$_{KP}$]$_{DP}$].}// \glb John cook-\textsc{(dir)-3sg.erg} \textsc{det} \textsc{loc} pot// \glft John cooked it in the pot.// %2594LL1[instrumental] \endgl \xe Because locative markers and the oblique marker are in complementary distribution, it is reasonable to assume that they occur in the same syntactic position. I label both oblique-marked nominals (\ref{1b}a) as well as nominals which form a constituent with a locative marker (\ref{1b}b) as KPs, since the oblique marker and the set of locative markers both carry case information, and designate a nominal as standing in an oblique grammatical relation to the main predicate.\footnote{The oblique marker signals that a nominal is a \emph{core oblique}, and a locative marker signals that a nominal is a \emph{non-core-oblique}, or locative adjunct in other words \citep[42-44]{Kroeber:1999}.}$^,$\footnote{\citet{BittnerHale:1996} posit KP as the nominal equivalent of CP in the verbal domain. They assume that K selects a DP for an argument, rather than D selecting a KP which is what I assume for Okanagan. My analysis is non-standard, since the relation between the selecting predicate head and the case-marked nominal is non-local (i.e. there is an intervening D-head). But despite being non-standard, some version of my analysis may be necessary, since it is undesirable to assume for (\ref{2b}) that \emph{i\textglotstop} is a K, or that \emph{t} is a D. There are 2 main points against this: First, such an analysis must analyze DP-internal locative markers as determiners, which ignores the fact that historically they were never determiners. Second, \emph{i\textglotstop} is semantically speaking a context-sensitive domain restrictor \citep{Lyon:2011}, a role argued by \citet{Gillon:2009b} to be universally associated with the D position. Also, under the assumption that D always selects for a KP, there must be a null case-marker for subject and transitive object DPs.} Evidence that the oblique marker and locative markers project their own syntactic category comes from conjunction \citep{Lyon:inprep}, and NP-deletion in relative clause formation. I assume the following basic DP structure for Okanagan: \ex\label{2b} \small{\Tree[.DP [.D {i\textglotstop}\\{\emph{the/a}} ] [.KP [.K {t} ] [.NP [.N tk\textbeltl m\'ilx\textsuperscript{w}\\{\emph{woman}} ] ] ] ]} \xe The distribution of the determiner \emph{i\textglotstop} and the oblique marker and locative markers across various grammatical categories is shown in the table below. \begin{center} \begin{tabular}{|l|c|c|c|}\hline & D & K & N\\ \hline {{Subjects}} & {{i\textglotstop}} & $\oslash$ & {tk\textbeltl m\'ilx\textsuperscript{w}}\\ {Transitive objects} & {{i\textglotstop}} & $\oslash$ & {tk\textbeltl m\'ilx\textsuperscript{w}}\\ {Oblique objects} & $\oslash$ & {t} & {tk\textbeltl m\'ilx\textsuperscript{w}}\\ {Applicative (\emph{-xt-}) Themes} & $\oslash$ & {t} & {tk\textbeltl m\'ilx\textsuperscript{w}}\\ {Passive agent obliques} & {{(i\textglotstop)}} & {t} & {tk\textbeltl m\'ilx\textsuperscript{w}}\\ {Instrumental obliques} & {(i\textglotstop)} & {t} & {tk\textbeltl m\'ilx\textsuperscript{w}}\\ {Locative adjuncts} & {(i\textglotstop)} & {\{\gx{k}l, l, tl\}} & {tk\textbeltl m\'ilx\textsuperscript{w}}\\ \hline \end{tabular} \begin{center}Table 1. \emph{Distribution of Nominal-Introducing Particles in Okanagan}\end{center} \end{center} This distribution provides important evidence for an analysis in which relative clauses in Okanagan are formed by clause-internal movement of a DP. \subsection{\hspace*{.04in}Relative Clauses} Relative clauses may modify an NP directly dominated by either a DP (\ref{1c}), or a KP (\ref{1d}). Additionally, the modifying clause may either precede or follow the head, as may be seen by comparing (\ref{1c}a) with (\ref{1c}b), and (\ref{1d}a) with (\ref{1d}b). \pex[labeloffset=.20in,textoffset=1.5ex]\label{1c} \a\begingl \gla {wa\g{y}} {\g{c}a\textglotstop-nt-\'is} {\textbf{i\textglotstop }} {sq\textschwa ltm\'ix\textsuperscript{w}} {\textbf{i\textglotstop }} {wik-s.}// \glb yes punch-\textsc{dir-3sg.erg} \textsc{det} man \textsc{det} see-\textsc{(dir)-3sg.erg}// \glft He hit the man he saw.// \endgl \a\begingl \gla {wa\g{y}} {\g{c}a\textglotstop-nt-\'is} {\textbf{i\textglotstop }} {wik-s} {\textbf{i\textglotstop }} {sq\textschwa ltm\'ix\textsuperscript{w}.}// \glb yes punch-\textsc{dir-3sg.erg} \textsc{det} see-\textsc{(dir)-3sg.erg} \textsc{det} man// \glft He hit the man he saw.//%2909LL,RS1[rel, preposed] \endgl \xe \pex~[labeloffset=.20in,textoffset=1.5ex]\label{1d} \a\begingl \gla {John} {\g{k}\textsuperscript{w}u\g{l}-\textschwa m} {\textbf{t}} {yam\v{x}\textsuperscript{w}a\textglotstop} {\textbf{t}} {k\textbeltl-s-n-\g{q}\textsuperscript{w}i\textbeltl-t\textschwa n-s.}// \glb John make-\textsc{mid} \textsc{obl} basket \textsc{obl} \textsc{unr.poss-nom}-n-pack-\textsc{instr-3sg.poss}// \glft John made the basket he was going to carry.// \endgl \a\begingl \gla {John} {\g{k}\textsuperscript{w}u\g{l}-\textschwa m} {\textbf{t}} {k\textbeltl-s-n-\g{q}\textsuperscript{w}i\textbeltl-t\textschwa n-s} {\textbf{t}} {yam\v{x}\textsuperscript{w}a\textglotstop .}// \glb John make-\textsc{mid} \textsc{obl} \textsc{unr.poss-nom}-n-pack-\textsc{instr-3sg.poss} \textsc{obl} basket// \glft John made the basket he was going to carry.// %3291LL1[rel, t rel, preposed, pass] \endgl \xe %CHECK. Is KP movement occurring here? I refer to head-initial relatives like (\ref{1c}a) and (\ref{1d}a) as \emph{post-nominal}, and head-final relatives like (\ref{1c}b) and (\ref{1d}b) as \emph{pre-posed}, following \citet{Davis:2010a}. Okanagan relatives must have particles (i.e. determiners and/or case markers) introducing both the head and the clausal remnant.\footnote{This effectively excludes \emph{pre-nominal} (DET [CLAUSE NP]) and \emph{post-posed} (DET [NP CLAUSE]) relatives as possibilities in Okanagan, although they are possible in other Salish languages, as shown in the following chart: \begin{center} \scriptsize{\begin{tabular}{|l||c|c|c|c|}\hline &Pre- & Post- & Post- & Pre-\\ &nominal & posed & nominal & posed\\ \hline\hline St'\'at'imcets & \checkmark & \checkmark & \checkmark & x\\ N\textbeltl\textschwa\textglotstop kepmxc\'in & x & x & \checkmark & \checkmark\\ Okanagan & x & x & {\checkmark} & {\checkmark}\\ Nxa\textglotstop amxcín& \checkmark& \checkmark& (\checkmark) & x\\ \hline \end{tabular}} \end{center} %CHECK how to prevent Moses pre-posed clauses? In Nxa\textglotstop amxcín, post-nominal relatives are possible \citep[124]{Mattina:2006}, but the oblique marker is becoming 'optional' here \citep[109]{Willett:2003}. I discuss my analysis of Nxa\textglotstop amxcín pre-nominal relatives in section 5.} Any grammatical role may be relativized in Okanagan. In addition to the relativized transitive and intransitive objects seen above in (\ref{1c}) and (\ref{1d}), subjects may be relativized (\ref{1e}) as well as themes of ditransitives (\ref{1f}): \ex[textoffset=.18in]\label{1e} \begingl \gla {\textbf{i\textglotstop }} {sq\textschwa ltm\'ix\textsuperscript{w}} {\textbf{i\textglotstop }} {\g{q}\textschwa \g{y}-nt-\'is} {i\textglotstop } {\g{q}\textschwa \g{y}min} {k\textsuperscript{w}u} {x\textsuperscript{w}i\g{c}-xt-s} {\textschwa nts\textglotstop a} {t} {i-k\textbeltl-\g{q}\textschwa \g{y}min.}// \glb \textsc{det} man \textsc{det} write-\textsc{dir-3sg.erg} \textsc{det} book \textsc{1sg.gen} give-\textsc{ditr-3sg.erg} \textsc{1sg.indep} \textsc{obl} \textsc{1sg.poss-unr.poss}-book// \glft The man who wrote the book gave me a book.// %2695LL1[rel] \endgl \xe \ex~[textoffset=.18in]\label{1f} \begingl \gla {k\textsuperscript{w}in-t} {\textbf{i\textglotstop }} {q\'aqx\textsuperscript{w}\textschwa lx} {\textbf{i\textglotstop }} {x\textsuperscript{w}i\g{c}-xt-m-n.}// \glb take-\textsc{dir} \textsc{det} fish \textsc{det} give-\textsc{ditr-2sg.acc-1sg.erg}// \glft Take the fish that I'm giving you.// %3268LL0[imper, postnominal rel] \endgl \xe % \begin{exe} % {\ex \gll {k\textschwa n} {\v{x}m\'ink\textschwa m} {\textbf{t}} {n\v{x}\textschwa s\'itk\textsuperscript{w}} {t} {si\g{w}\textbeltl k\textsuperscript{w}} {\textbf{t}} {ks-siwst-x} {i-s\g{l}\'a\v{x}t.}\\ \textsc{1sg.abs} want-\textsc{mid} \textsc{obl} {n-good-water} \textsc{obl} water \textsc{obl} \textsc{fut}-drink-\textsc{incept} \textsc{1sg.poss}-friend\\ I need water that my friend can drink.} %3266LL1[t rel] % \label{x1} Unlike many other Salish languages, Okanagan relative clauses do not exhibit any special inflectional pattern. In other words, pronominal morphology found on relative clauses may also be found in main clauses \citep{Kroeber:1999}.\footnote{There are nevertheless differences in the distribution of main versus subordinate clause inflectional patterns. It is difficult to extract an intransitive oblique object from a predicate inflected with the \emph{-m} or \emph{(-m\'ixa\textglotstop)x} intransitive suffixes, although apparently possible with a third person subject: \ex[exno=i] \begingl \gla {kn} {\v{x}m\'ink-\textschwa m} {t} {siw\textbeltl k\textsuperscript{w}} {t} {ks-s\'iwst-x} {i-s\g{l}\'a\v{x}t}.// \glb \textsc{1sg.abs} want-\textsc{mid} \textsc{obl} water \textsc{obl} \textsc{fut}-drink-\textsc{intr} \textsc{1sg.poss}-friend// \glft I want some water for my friend to drink.// \endgl \xe On the other side of the coin, nominalized possessor predicates, such as \emph{i-sc-w\'ik} ``my seeing''/``I saw'' in (\ref{1f}b), are generally not used as main-clause predicates, although nominalized future forms inflected with a middle suffix, such as \emph{i-ks-p\'ulst\textschwa m} ``I'm going to beat him'' are often found in non-embedded contexts.} (\ref{1f}a) shows a pre-posed object relative inflected with an ergative subject, and (\ref{1f}b) shows the corresponding nominalized object relative with a possessor subject.\footnote{The exact semantic difference between (\ref{1f}a) and (\ref{1f}b), if there actually is one, remains unclear. Speakers indicate that nominalized forms like (\ref{1f}b) are past-tense completive, while ergative forms like (\ref{1f}a) are present-tense completive, but my research suggests that there is no clear demarcation between the two, and that both can be uttered felicitously within an identical discourse situation. Nominalized relatives may have less clausal structure than relatives inflected with ergative subjects \citep{Thompson:2011}, but since nominalized clauses can function as main predicates in Okanagan and select for DP arguments, the case can also be made that extraction of such an argument from a nominalized clause involves clause-internal movement. On that note, it is not yet established whether there is any difference between predicate and clausal nominalization in subordinate clause contexts, since there are no pre-predicative auxiliaries in Okanagan to which a nominalizer might attach, thereby providing evidence for a distinction betwen predicate and clausal nominalization.}$^,$\footnote{The determiner \emph{i\textglotstop} regularly elides before 1st person possessive prefix \emph{in-} and 2nd person possessive prefix \emph{an-}, as in (\ref{1f}b), and lowers to \emph{a\textglotstop} before customary prefix \emph{(a)c-}, as in (\ref{1h}b).} \pex[labeloffset=.20in,textoffset=1.5ex]\label{1f} \a\begingl \gla {\g{t}\textrevglotstop \'ap-nt-\'in} {\textbf{i\textglotstop }} {w\'ik-\textschwa n} {\textbf{i\textglotstop }} {sk\textschwa k\textrevglotstop\'aka\textglotstop .}// \glb shoot-\textsc{dir-1sg.erg} \textsc{det} see-\textsc{(dir)-1sg.erg} \textsc{det} bird// \glft I shot the bird that I've seen.//% (LL, VG) %6245LL1[cleft, rel] Just about the same thing. \endgl \a\begingl \gla {\g{t}\textrevglotstop \'ap-nt-\'in} {\textbf{(i\textglotstop)}} {i-sc-w\'ik} {\textbf{i\textglotstop }} {sk\textschwa k\textrevglotstop\'aka\textglotstop .}// \glb shoot-\textsc{dir-1sg.erg} \textsc{(det)} \textsc{1sg.poss-perf}-see \textsc{det} bird// \glft I shot the bird that I've seen.// %(LL, VG) %6244LL1[cleft, rel] nominalized clause preposed? \endgl \xe%HD: what happens in a preposed relative with third person transitive subject morphology? Is it still interpreted as a headed relative, or would it be a headless relative with a DP object (e.g. 'the one that saw the bird')? Or would it be ambiguous. When transitive subjects are relativized, speakers often prefer to passivize the predicate. In (\ref{1g}), the clausal remnant is inflected as passive by the suffix \emph{-m}, and is introduced by the sequence \emph{i\textglotstop} \emph{t}, which together indicate that the passive agent has been extracted:\footnote{See example (\ref{rca}) for a similar case.} \ex[textoffset=.10in]\label{1g} \begingl \gla {sc-\g{\textcrlambda}a\textglotstop \g{\textcrlambda}a\textglotstop-\'am-s} {\textbf{i\textglotstop }} {s\textschwa x\textsuperscript{w}\g{m}a\textglotstop\g{m}\'aya\textglotstop m-s} {\textbf{i\textglotstop }} {\textbf{t}} {knx\'it-(t)-m} {i\textglotstop } {l} {s\textschwa n\g{q}\textschwa \g{y}m\'int\textschwa n.}// \glb \textsc{impf}-look.for-\textsc{mid-3sg.poss} \textsc{det} teacher-\textsc{3sg.poss} \textsc{det} \textsc{obl} help-\textsc{dir-pass} \textsc{det} \textsc{loc} school// \glft He's looking for the teacher that helped him at school.// \endgl \xe Headless relatives are also common in Okanagan (\ref{1h}). I assume that these are a special type of post-nominal relative, where the head noun, and its selecting determiner, are both null.\footnote{\citet{Davis:2010a} argues against a similar analysis for St'át'imcets, instead claiming that relatives in this language are all derived from a common pre-nominal structure. His analysis will not work for Okanagan, however, since Okanagan (unlike St'\'at'imcets) has pre-posed relatives.} \pex[labeloffset=.13in,textoffset=1.5ex]\label{1h} \a\begingl \gla {\g{q}\textschwa \g{y}nt\'ix\textsuperscript{w}} {i\textglotstop } {q\textsuperscript{w}\textschwa lq\textsuperscript{w}\'ilstm\textschwa n.}// \glb write-\textsc{dir-2sg.erg} \textsc{det} speak-\textsc{caus-2sg.abs-1sg.erg}// \glft Write down what I'm telling you.//% (LL).%1918LL1[dem][rel] [cf GW:49] \endgl \a\begingl \gla {ka\textglotstop k\'ic\textschwa n} ({i\textglotstop}) {acs\g{l}m\'ist\textschwa n.}// \glb find-\textsc{(dir)-1sg.erg} \textsc{(det)} \textsc{cust}-lose-\textsc{caus-1sg.erg}// \glft I found the one I was looking for.// \endgl \xe Demonstratives appear to function as relative clause heads (\ref{1i}), but since demonstratives often adjoin to a constituent DP \citep{Lyon:2010a}, (\ref{1i}) may also be analyzed as a headless relative if we assume that the adjoined DP is null in these cases. \ex[textoffset=.10in]\label{1i} \begingl \gla {w\'ik-\textschwa n} {ix\'i\textglotstop} {i\textglotstop} {ks-knx\'it-m-s.}// \glb see-\textsc{(dir)-1sg.erg} \textsc{dem} \textsc{det} \textsc{fut}-help-\textsc{dir-2sg.obj-3sg.erg}// \glft I saw the one who will help you.// \endgl \xe I now move on to a more technical discussion of the syntactic processes involved in relative clause formation in Okanagan. %CHECK Head must be a noun. \section{\hspace*{.15in}Relative clause formation by movement} As first noted by \citep[396]{Kroeber:1997} for N\textbeltl\textschwa\textglotstop kepmxc\'in, locative relative clauses seem to involve clause internal movement of a DP to the left periphery of a relative clause. Kroeber notes that in examples like (\ref{1}), ``...the preposition codes the relation of gap to relative clause predicate, not the relation of the whole relative clause to the matrix predicate.'' \ex[textoffset=.10in]\label{1} \begingl \gla {(w)\textglotstop\'ex} {kn} {x\textsuperscript{w}\'i\textglotstop-m} {te} {np\'uytn$_2$} { } [[{\textbf{n-e}} [${\oslash}_{NP_2}$]$_{PP_1}$] {x\textsuperscript{w}\'u\g{y}} {wn} {\textrevglotstop\textsuperscript{w}\'o\g{y}t} {\emph{$t_1$}$_{CP}$]}// \glb \textsc{prog} \textsc{1sg} look.for-\textsc{mid} \textsc{obl.det} bed { } in-\textsc{det} { } \textsc{fut} \textsc{1sg.conj} sleep { }// \glft I'm looking for a bed where I'm gonna' sleep. \citep[132]{Koch:2006}// \endgl \xe In other words, because the preposition \emph{n} ``in'' in (\ref{1}) helps to specify the location of the sleeping event, and not the looking event, the preposition may plausibly be analyzed as having moved from a base position following the relative clause. \citet{Davis:2004} and \citet{Koch:2006} have shown for St'\'at'imcets and N\textbeltl\textschwa\textglotstop kepmxc\'in respectively, that the determiner also moves, or rather, the DP ``pied-pipes'' the preposition to a clause-initial position. This is illustrated by the bracketing in (\ref{1}).\footnote{Since N\textbeltl\textschwa\textglotstop kepmxc\'in determiners vary with regards to their spatio-temporal properties, \citet{Koch:2006} is able to show that the determiner introducing the relative clause shows the spatio-temporal properties of the relative clause predicate, rather than the main clause predicate, confirming that movement also occurs in relatives which do not involve locative marking. For Okanagan, it is not possible to use different determiners as a diagnostic for movement, since there is only one determiner involved in relativization, \emph{i\textglotstop}. Nevertheless, the oblique marker \emph{t} as well as the other locative markers, help to confirm that movement has occurred.} Recall that for Okanagan, the oblique marker \emph{t} and locative markers \emph{\gx{k}l, l} and \emph{tl} may co-occur with \emph{i\textglotstop}. These particle sequences help provide evidence for clause-internal movement. In main clauses, the combination of \emph{i\textglotstop} and \emph{t} introduces instruments and passive agents, as in (\ref{2}), and the combination of \emph{i\textglotstop} and a locative particle designates a DP as a locative adjunct, as in (\ref{4}) \emph{{i\textglotstop } {tl} {sq\textschwa ltm\'ix\textsuperscript{w}}} ``from the man''. \pex[labeloffset=.13in,textoffset=1.5ex]\label{2} \a\begingl \gla {\g{t}\textrevglotstop ap-nt-\'is} [\textbf{i\textglotstop } [\textbf{t} {swlwlm\'ink}$_{KP}$]$_{DP}$].// \glb shoot-\textsc{dir-3sg.erg} \textsc{det} \textsc{obl} gun// \glft He shot it with a gun.// \endgl \a\begingl \gla {Mike} {\g{c}\'u\g{m}qs-nt-m} [\textbf{i\textglotstop } [\textbf{t} {tk\textbeltl m\'ilx\textsuperscript{w}}$_{KP}$]$_{DP}$].// \glb Mike kiss-\textsc{dir-pass} \textsc{det} \textsc{obl} woman// \glft Mike was kissed by the woman.// \endgl \xe \ex~[textoffset=.10in]\label{4} \begingl \gla {{ac-ylt-m\'i-st-l\textschwa x}} [\textbf{i\textglotstop } [\textbf{tl} {sq\textschwa ltm\'ix\textsuperscript{w}}$_{KP}$]$_{DP}$].// \glb \textsc{cust}-{run.away}-\textsc{appl-caus-3pl.erg} \textsc{det} \textsc{loc} man// \glft They're running away from the man.// \endgl \xe In support of a movement analysis for Okanagan relatives, consider that when instruments and passive agents like those in (\ref{2}) are relativized, the relative clause is introduced by both \emph{i\textglotstop} and \emph{t}, as in (\ref{3}): \pex[labeloffset=.13in,textoffset=1.5ex]\label{3} \a\begingl \gla {k\textsuperscript{w}u} {\textbeltl i\textglotstop\g{q}\textsuperscript{w}-m-\textbeltl t} {i\textglotstop } {ni\g{k}-mn$_{2}$} [[\textbf{i\textglotstop} \textbf{t} [\ensuremath{\oslash}$_{NP_2}$]$_{DP_1}$] {ni\g{k}-nt-x\textsuperscript{w}} \emph{t$_1$}$_{CP}$].// \glb \textsc{1sg.gen} show-\textsc{ditr} \textsc{det} knife \textsc{det} \textsc{obl} { } cut-\textsc{dir-2sg.erg}// \glft Show me the knife that you cut it with.// %6669LL1[rel, instr] \endgl \a\begingl \gla {Mike} {wiks} {i\textglotstop } {tk\textbeltl m\'ilx\textsuperscript{w}$_2$} [[\textbf{i\textglotstop } \textbf{t} [\ensuremath{\oslash}$_{NP_2}$]$_{DP_1}$] {\g{c}\'um\g{q}s-nt-m} \emph{t$_1$}$_{CP}$].// \glb Mike see-\textsc{(dir)-3sg.erg} \textsc{det} woman \textsc{det} \textsc{obl} { } kiss-\textsc{dir-pass}// \glft Mike saw the woman he was kissed by.// %4818LL1Co-valuation [condition C, pro, object relative] \endgl \xe Note that \emph{i\textglotstop} and \emph{t} normally only co-occur when introducing a passive agent or instrument, or before clauses from which these grammatical roles have been extracted. In extraction contexts involving passive patients, for example, \emph{i\textglotstop} \emph{t} may not introduce the relative clause, only \emph{i\textglotstop}. (\ref{3a}) shows an example of an extracted patient, where the clausal remnant is introduced by the determiner \emph{i\textglotstop}, and an in-situ clause-internal agent is introduced by \emph{i\textglotstop} \emph{t}. \ex[textoffset=.10in]\label{3a} \begingl \gla {John} {s\textschwa c\g{\textcrlambda}a\textglotstop \g{\textcrlambda}a\textglotstop \'ams} \textbf{i\textglotstop } {tk\textbeltl milx\textsuperscript{w}} {i\textglotstop } \textbf{(*t)} {knx\'it\textschwa m} {i\textglotstop } {t} {sq\textschwa ltm\'ix\textsuperscript{w}.}// \glb John \textsc{impf}-{look for}-\textsc{mid}-\textsc{3sg.poss} \textsc{det} woman \textsc{det} \textsc{(*obl)} help-\textsc{(dir)-pass} \textsc{det} \textsc{obl} man// \glft John is looking for the woman who was helped by the man.// \endgl \xe Given that the distribution of the sequence \emph{i\textglotstop} \emph{t} is limited to the same grammatical subset in both extraction and non-extraction contexts, the sequence \emph{i\textglotstop} \emph{t} in (\ref{3}) constitutes evidence for clause-internal movement. Similarly, when a locative adjunct is extracted in Okanagan, the relative clause is introduced by a determiner plus locative marker sequence, thus furnishing evidence parallel to N\textbeltl\textschwa\textglotstop kepmxc\'in (\ref{1}) that clause-internal movement has indeed occurred. Compare (\ref{4}) and (\ref{4b}a), in particular. \pex[labeloffset=.13in,textoffset=1.5ex]\label{4b} \a\begingl \gla {wik-\textschwa n} {i\textglotstop } {sq\textschwa ltm\'ix\textsuperscript{w}$_2$} {[[\textbf{i\textglotstop} } \textbf{tl} [\ensuremath{\oslash}$_{NP_2}$]$_{DP_1}$] {{ac-ylt-m\'i-st-l\textschwa x}} \emph{t$_1$}$_{CP}$].// \glb see-\textsc{(dir)-1sg.erg} \textsc{det} man \textsc{det} \textsc{loc} { } \textsc{cust}-{run.away}-\textsc{appl-caus-3pl.erg} { }// \glft I see the man that they're running away from.// %4114LL1[rel, loc] \endgl \a\begingl \gla {uc} {c-my-st-\'ix\textsuperscript{w}} {i\textglotstop} {sq\textschwa ltm\'ix\textsuperscript{w}} \textbf{i\textglotstop} \textbf{\g{k}l} {tw-m\'i-st-\textschwa m-\textschwa n} {i\textglotstop} {lasm\'ist}.// \glb \textsc{ynq} \textsc{cust}-know-\textsc{caus-2sg.erg} \textsc{det} man \textsc{det} \textsc{loc} sell-\textsc{appl-caus-appl(?)-1sg.erg} \textsc{det} shirt// \glft Do you know the man that I sold the shirt to?// %LL read back, good \endgl \xe Following \citet{Davis:2010a} and \citet{Koch:2006}, it seems clear that for Okanagan (\ref{3}) and (\ref{4b}) at least, a DP internal to the relative clause has raised to the left periphery of the relative clause CP. The noun in the moved DP then plausibly undergoes deletion through identity with the clause exterior head NP. The following structure is thus a plausible representation of the relative clause in (\ref{4b}a): \ex\label{4c} \scriptsize{\Tree[.DP [.D {i\textglotstop} ] [.KP [.K \ensuremath{\oslash} ] [.NP [.NP [.N {sq\textschwa ltm\'ix\textsuperscript{w}_j} ] ] [.CP [.Spec [.DP_i [.D \textbf{i\textglotstop} ] [.KP [.K \textbf{tl} ] [.NP {pro_j} ] ] ] ] [.C' [.C {\ensuremath{\oslash}} ] [.TP ... [.VP {acylt\textschwa m\'istl\textschwa x} [.DP\emph{t_i} ] ] ] ] ] ] ] ]} \xe Assuming that \emph{all} relative clauses in Okanagan are similarly formed, the structure in (\ref{4c}) implies that the sequence of particles introducing the clausal remnant should \emph{always} code the relation of the gap to the relative clause predicate. This holds true in some cases. For example, consider that subject and transitive object extractions in Okanagan, e.g. (\ref{1c}) and (\ref{1e}), are characterized by having the determiner \emph{i\textglotstop} introduce both the head and the clausal remnant. Since transitive predicates always select for \emph{i\textglotstop} DP objects in main clause contexts (\ref{4d}a), the prediction is that when an object is extracted, the clausal remnant will be introduced by only a determiner \emph{i\textglotstop}. This prediction is upheld (\ref{4d}b).\footnote{But admittedly these constitute only weak support for movement, since as \citet{Koch:2006} notes for similar cases in N\textbeltl\textschwa\textglotstop kepmxc\'in, the two determiners may simply be copies of one another.} \pex[labeloffset=.13in,textoffset=1.5ex]\label{4d} \a\begingl \gla {wik-s} {i\textglotstop} {sq\textschwa ltm\'ix\textsuperscript{w}}// \glb see-\textsc{(dir)-3sg.erg} man// \glft He saw the man.// \endgl \a\begingl \gla {wa\g{y}} {\g{c}a\textglotstop-nt-\'is} {i\textglotstop } {sq\textschwa ltm\'ix\textsuperscript{w}} \textbf{i\textglotstop } {wik-s}.// \glb yes punch-\textsc{dir-3sg.erg} \textsc{det} man \textsc{det} see-\textsc{(dir)-3sg.erg}// \glft He hit the man he saw.// \endgl \xe As a working hypothesis then, I assume that \emph{all} Okanagan relatives are formed by clause-internal movement. Relative clauses are canonically post-nominal, and pre-posed relatives are derived from post-nominals by an additional movement of the relative clause CP to a position preceding the DP containing the head, presumably Spec DP.\footnote{Or possibly adjoined to DP. Pre-posed relatives in Okanagan are generally more marked than post-nominal forms, for reasons to be discussed.} Compare the post-nominal relative clause \emph{i\textglotstop { }sq\textschwa ltm\'ix\textsuperscript{w} i\textglotstop { }k\textsuperscript{w}u wiks} ``the man who saw me'' (\ref{4e}a) with its equivalent pre-posed version \emph{i\textglotstop { }k\textsuperscript{w}u wiks{ }i\textglotstop { }sq\textschwa ltm\'ix\textsuperscript{w}} (\ref{4e}b): %HD: have you tried to stack relative clauses? Does a postnominal vs preposed asymmetry appear when you do? And suppose you have long-range movement in the relative clause: is the preposed structure possible then? \pex[labeloffset=.13in]\label{4e} \a{\scriptsize{\Tree[.DP [.D {i\textglotstop} ] [.KP [.K {\ensuremath{\oslash}} ] [.NP [.NP [.N {sq\textschwa ltm\'ix\textsuperscript{w}_j} ] ] [.CP [.Spec [.DPt_i [.D {i\textglotstop} ] [.KP [.K {\ensuremath{\oslash}} ] [.NP [.N {pro_j} ] ] ] ] ] [.C' [.C {\ensuremath{\oslash}} ] [.TP ... [.VP {k\textsuperscript{w}u wiks} [.DP_i ] ] ] ] ] ] ] ]}} \a{\scriptsize{\Tree[.DP [.Spec [.CP_k [.Spec [.DPt_i [.D {i\textglotstop} ] [.KP [.K {\ensuremath{\oslash}} ] [.NP [.N {pro_j} ] ] ] ] ] [.C' [.C \ensuremath{\oslash} ] [.TP ... [.VP {k\textsuperscript{w}u wiks} [.DP_i ] ] ] ] ] ] [.D' [.D {i\textglotstop} ] [.KP [.K {\ensuremath{\oslash}} ] [.NP [.NP [.N {sq\textschwa ltm\'ix\textsuperscript{w}_j} ] ] [.CPt_k ] ] ] ] ]}} \xe Extending this movement account to all Okanagan relatives encounters several problems, however. First and foremost, the distribution of \emph{t} before a clause does not always code the relation of the gap to the relative clause predicate.\footnote{Other issues which require further investigation are: (i) Whether KP movement (rather than DP movement) occurs for cases where a relative clauses modifies a KP-contained head (\ref{1d}), or whether there may be a null determiner in the language; (ii) An explanation for the `matching effect' displayed between the head-introducing and clause-introducing particles, a phenomenon which I touch on in the next section; (iii) Ditransitive theme extractions, which still do not follow from my extension of the movement account.} The next section discusses these problematic data. \section{\hspace*{.15in}Extending the movement account} \subsection{\hspace*{.04in}Problematic cases of oblique marking} Not all relative clauses in Okanagan conform so nicely to the movement account described in the previous section. Consider that relative clause predicates may be inflected with the future prefix \emph{ks-}. In these cases, the clausal remnant is often introduced by both \emph{i\textglotstop} and \emph{t}, but this sequence does \emph{not} code the relation of the gap to the relative clause predicate. (\ref{5a}) shows that in main clause contexts, a future transitive predicate cannot select for an object introduced by \emph{i\textglotstop} \emph{t}\footnote{Recall that for (\ref{5a}), an oblique marker is not possible, since `the drum' is a grammatical object, and not an instrument or passive agent.}, yet in extraction contexts (\ref{5b}), the oblique marker \emph{t} may co-occur with the determiner.\footnote{From this point onwards, I highlight the oblique marker which introduces a relative clause in blue type, to help the reader distinguish between this occurence of \emph{t}, and its other role as a nominal case marker.} \ex[textoffset=.10in]\label{5a} \begingl \gla {ks-ya\textglotstop-y\'a\textglotstop \v{x}a\textglotstop-s\textschwa lx} {\textbf{i\textglotstop }} {(*{\textbf{t}})} {pwm\'in}// \glb \textsc{fut}-show-\textsc{(dir)-3pl.erg} \textsc{det} \textsc{(*obl)} drum// \glft They will look at a drum.// \endgl \xe \ex~[textoffset=.10in]\label{5b} \begingl \gla {wa\g{y}} {i-ks-\g{k}\textsuperscript{w}u\g{l}-\textschwa m} {\textbf{i\textglotstop }} {pwm\'in} {\textbf{i\textglotstop }} {\textbf{({\color{blue}{t}})}} {ks-ya\textglotstop-y\'a\textglotstop \v{x}a\textglotstop-s\textschwa lx.}// \glb yes \textsc{1sg.poss-fut}-make-\textsc{mid} \textsc{det} drum \textsc{det} \textsc{(obl)} \textsc{fut}-show-\textsc{(dir)-3pl.erg}// \glft I will make a drum that they will look at.// %the nominalized matrix clause is confusing here: would you get the same effect with an ordinary (formally transitive) matrix verb? %4127LL1[rel, irrealis, temp] % \item[(c)] \gll *{wa\g{y}} {iks\g{k}\textsuperscript{w}u\g{l}\textschwa m} {\textbf{i\textglotstop }} {pewminn} {\textbf{t}} {ksya\textglotstop y\'a\textglotstop \v{x}a\textglotstop s\textschwa lx.}\\ yes \textsc{1sg.poss-fut}-make-\textsc{mid} \textsc{det} drum \textsc{obl} \textsc{fut}-show-\textsc{(dir)-3pl.erg}\\ I will make a drum that they will look at. %4127LL1[rel, irrealis, temp] \endgl \xe (\ref{6a}-\ref{6b}) show that the same pattern surfaces with subjects. In main clause contexts, subjects of future-inflected transitives may not be introduced by the oblique marker \emph{t}, only by \emph{i\textglotstop}, but (\ref{6b}a,b) confirms that in extraction contexts, the sequence \emph{i\textglotstop} \emph{t} is possible. \ex[textoffset=.10in]\label{6a} \begingl \gla {ks-knx\'it-m-s} {\textbf{i\textglotstop }} {(*{\textbf{t}})} {sq\textschwa ltm\'ix\textsuperscript{w}}// \glb \textsc{fut}-help-\textsc{(dir)-2sg.acc-3sg.erg} \textsc{det} \textsc{(*obl)} man// \glft The man will help you.// \endgl \xe \pex~[labeloffset=.13in,textoffset=1.5ex]\label{6b} \a\begingl \gla {kn} {wik\textschwa m} {\textbf{t}} {sq\textschwa ltm\'ix\textsuperscript{w}} {\textbf{(i\textglotstop)}} {\textbf{\color{blue}{t}}} {ks-knx\'it-m-s.}// \glb \textsc{1sg.abs} see-\textsc{mid} \textsc{obl} man \textsc{det} \textsc{obl} \textsc{fut}-help-\textsc{(dir)-2sg.obj-3sg.erg}// \glft I saw a man who will help you.// %4322LL,NS0[det, rel] \endgl \a\begingl \gla {w\'ik-\textschwa n} {\textbf{i\textglotstop }} {sq\textschwa ltm\'ix\textsuperscript{w}} {\textbf{i\textglotstop }} ({\textbf{\color{blue}{t}}}) {ks-knx\'it-m-s.}// \glb see-\textsc{(dir)-1sg.erg} \textsc{det} man \textsc{det} \textsc{obl} \textsc{fut}-help-\textsc{(dir)-2sg.acc-1sg.erg}// \glft I saw the man that will help you.// %4558LL1[det, rel, mod, trans] \endgl \xe At this point, it is worthwhile to briefly discuss the apparent `matching' relation which holds between the head-introducing and clause-introducing particles in (\ref{6b}a) and (\ref{6b}b). Notice that for (\ref{6b}b), the oblique marker \emph{t} is optional, while for (\ref{6b}a), it is the determiner \emph{i\textglotstop} that is optional. This difference appears to be due to the transitivity of the main clause predicate. Recall that in Okanagan, formally transitive predicates like \emph{w\'ik\textschwa n} ``I see s.t.'' in (\ref{6b}b) will always select a full DP as an object, while formally intransitive predicates like \emph{wik\textschwa m} ``to see'' (\ref{6b}a) will always select for an oblique-marked KP as an object. This tight correlation between predicate transitivity and nominal marking drives the matching relation between the particle introducing the head noun and which particle \emph{must} introduce the future-marked clausal remnant. The presence or absence of the optional particle in these cases seems to be a surface-level phenomena, there being no semantic difference between forms with and without the optional particle, and so it seems reasonable to assume that both particles are underlyingly present in these cases.\footnote{The fact that this `optional' oblique marker \emph{t} is only apparent in extractions from future-inflected predicates is interesting, especially since what I analyze as the equivalent particle in Nxa\textglotstop amxcín is not dependent on the tense/aspect properties of the clausal remnant. It is quite possible that this \emph{t} has been borrowed by the Upper Nicola dialect from N\textbeltl\textschwa\textglotstop kepmxc\'in, and that it occurs only before future \emph{ks-} on analogy with the N\textbeltl\textschwa\textglotstop kepmxc\'in \emph{tk} and Secwepemcts\'in \emph{tek} 'oblique+irrealis determiner'. In both N\textbeltl\textschwa\textglotstop kepmxc\'in and Secwepemcts\'in, this sequence occurs before `unrealized' or `irrealis' intransitive objects, and in N\textbeltl\textschwa\textglotstop kepmxc\'in at least, also before relative clause predicates with unrealized/irrealis heads. If this hypothesis is correct, the prediction is that this \emph{t} will be absent from other Okanagan dialects not so heavily influenced by N\textbeltl\textschwa\textglotstop kepmxc\'in. This point also brings to mind the historical connection between the irrealis \emph{k}-type determiners of the Northern Interior, and future \emph{ks-} in Okanagan, and raises the question of whether the historical developments discussed in this paper might not be related to the absence of a \emph{k} determiner in Okanagan, or its probable reanalysis as an aspectual prefix.} But here we encounter a problem. Neither (\ref{6b}a) nor (\ref{6b}b) are consistent with the movement account, since in neither case does the sequence \emph{i\textglotstop} \emph{t} code the relation of the gap to the clausal remnant. In other words, under the movement analysis, the prediction is that for both (\ref{6b}a) and (\ref{6b}b), \emph{i\textglotstop} must introduce the clausal remnant and \emph{t} should be ungrammatical, since these are not instrument or passive agent extractions. If the oblique marker \emph{t} did not undergo raising with a constituent clause-internal DP in these cases, then what is the function of \emph{t} here, and where did it come from? To begin to answer this question, it is important to note that Okanagan does show evidence that in certain cases, the oblique marker \emph{cannot} have moved as a constituent with a post-clausal DP. Examples (\ref{7a}a,b) below are the structural equivalents to examples (\ref{6b}a,b) above, the difference being that the pronominal object of the clausal remnant is realized as a pro-clitic, rather than a suffix. In these cases, \emph{t} occurs \emph{between} the object clitic and the remnant predicate:\footnote{See A. \citet{Mattina:1993} for a discussion of Okanagan pronominal paradigms.} \pex[labeloffset=.13in,textoffset=1.5ex]\label{7a} \a\begingl \gla {kn} {ks-\g{\textcrlambda}a\textglotstop \g{\textcrlambda}a\textglotstop-m\'ixa\textglotstop x} {\textbf{t}} {t\textschwa twit$_2$} [[{\textbf{(i\textglotstop)}} [${\oslash}_{NP_2}$]$_{DP_1}$] {k\textsuperscript{w}u} {\textbf{\color{blue}{t}}} {ks-knx\'it-s} {\emph{$t_1$}} {\textbeltl a\textglotstop } {\v{x}lap}$_{CP}$].// \glb \textsc{1sg.abs} \textsc{fut}-look.for-\textsc{incept} \textsc{obl} boy \textsc{(det)} { } \textsc{1sg.gen} \textsc{obl} \textsc{fut}-help-\textsc{(dir)-3sg.erg} { } \textsc{comp} tomorrow// \glft I'm gonna look for a boy to help me tomorrow.// %(LL, VG) %7260LL1[rel, det, obl] what\u2019s going on here with this rel?? \endgl \a\begingl \gla {\textglotstop aws-\g{\textcrlambda}a\textglotstop \g{\textcrlambda}a\textglotstop-nt-\'in} {\textbf{i\textglotstop }} {t\textschwa twit$_2$} [[{\textbf{i\textglotstop }} {[${\oslash}_{NP_2}$]$_{DP_1}$]} {k\textsuperscript{w}u} {\textbf{{({\color{blue}{t}})}}} {ks-knx\'it-s} {\emph{$t_1$}} {\textbeltl a\textglotstop } {\v{x}lap}$_{CP}$].// \glb go-look.for-\textsc{dir-1sg.erg} \textsc{det} boy \textsc{det} { } \textsc{1sg.gen} \textsc{(obl)} \textsc{fut}-help-\textsc{(dir)-3sg.erg} { } \textsc{comp} tomorrow// \glft I went looking and I found the boy who's gonna help me tomorrow.// %7262LL1[rel, det, obl] \endgl \xe % \ex \gll {\textglotstop aws\g{\textcrlambda}a\textglotstop \g{\textcrlambda}a\textglotstop nt\'in} {i\textglotstop } {t\textschwa twit} {i\textglotstop } {k\textsuperscript{w}u} {t} {ksknx\'its} {\textbeltl a\textglotstop } {\v{x}lap.}\\ \\ I went looking and I found the boy who's gonna help me tomorrow. %7263LL0[rel, det, obl] Since the 1st person object clitic \emph{k\textsuperscript{w}u} in (\ref{7a}a,b) is certainly not a constituent with the moved DP, it can safely be concluded that the oblique marker \emph{t}, which follows the clitic in this case, is also not a constituent with the moved DP, ergo it does not undergo movement. Concerning the position of \emph{k\textsuperscript{w}u}, I assume a morpho-phonological analysis of pronominal pro-clitics, whereby they attach to the left-most element of a clause. Since \emph{t} does not move in (\ref{7a}a,b), but rather delimits the left-periphery of the clause, a pronominal pro-clitic will attach to the left of \emph{t}.\footnote{This predicts that a DP object, whose position is \emph{not} morpho-phonologically determined, but syntactically determined, may not substitute for the object proclitic in such cases. While overt clause-internal nominal DP objects of subject-extracted relatives must occur after the clausal remnant, I have not yet tried to substitute an object DP for the proclitic in the cases shown above.} Note that pronominal pro-clitics also apparently attach to \emph{t} in cleft contexts (\ref{7h}a), and regularly precede the complementizer \emph{\textbeltl a\textglotstop} in contexts involving clausal subordination (\ref{7h}b):%While overt clause-internal nominal DP objects of subject-extracted relatives must occur after the clausal remnant, I have not yet tried to substitute an object DP for the proclitic in the cases shown above. \pex[labeloffset=.13in,textoffset=1.5ex]\label{7h} \a\begingl \gla {anw\'i\textglotstop} {k\textsuperscript{w}} {t} {sq\textschwa ltm\'ix\textsuperscript{w}} {i\textglotstop} {k\textsuperscript{w}} {ylm\'ix\textsuperscript{w}\textschwa m}.// \glb \textsc{2sg.indep} \textsc{2sg.abs} \textsc{obl} man \textsc{det} \textsc{2sg.abs} chief// \glft `You’re the man who is the chief.'// \endgl \a\begingl \gla {cak\textsuperscript{w}} {\v{x}ast} {k\textsuperscript{w}} {\textbeltl a\textglotstop} {ka\textglotstop k\'ic-\textschwa m} {t} {siw\textbeltl k\textsuperscript{w}}.// \glb \textsc{deon} good \textsc{2sg.abs} \textsc{comp} find-\textsc{mid} \textsc{obl} water// \glft `It'd be good if you go find some water.'// %(8039) \endgl \xe The implication from data like (\ref{7a}) is that when oblique \emph{t} precedes a relative clause, it does not necessarily code the relation of the gap to the clausal remnant, since it does not necessarily undergo movement. In other words, when a sequence \emph{i\textglotstop} \emph{t} precedes a relative clause, \emph{t} is either a case marker which moves as a constituent with a clause-internal DP (i.e. in instrument and passive agent extractions), or \emph{t} is something else (i.e. in argument extractions from future-marked predicates). There is a syntactic difference between these two different types of oblique marking. %HD: Notice that t actually doesn't precede the relative clause, either, since it follows the genitive subject in (26), whereas the D2 precedes it. TALK ABOUT MORPHO-phonology! I claim that the \emph{t} found in data like (\ref{5b}), (\ref{6b}), and (\ref{7a}) is a remnant of an earlier relativization strategy in Okanagan, whereby \emph{t} introduced all relative clauses. This claim is supported from data in neighboring Salish languages. Consider that in the Northern Interior Salish languages of N\textbeltl\textschwa\textglotstop kepmxc\'in \citep{Koch:2006} and Secwepemcts\'in \citep{Gardiner:1993}, non-locative clausal remnants are introduced by the oblique marker \emph{t(e)}, regardless of the grammatical status of the moved constituent. The Southern Interior language Nxa\textglotstop amxcín also exhibits data showing that its oblique marker \emph{t} cannot have moved together with a clause-internal DP. For these languages, as with Okanagan, only specific grammatical roles may be marked as oblique, and so assuming that all Interior Salish languages form relative clauses by a movement of a clause-internal DP to the left periphery of the clause, the relatively unrestricted occurrence of the oblique marker before relative clauses in these languages may be construed as evidence that it does not necessarily move with the clause-internal DP. In addition, an interesting and relevant difference emerges between Northern and Southern Interior Salish languages with regards to the linear position of the oblique marker in these cases, which in turn affects the mechanics of the movement analysis. It is to these languages that I now turn. %\footnote{It is also possible that \emph{t} in this case is a borrowing from N\textbeltl\textschwa\textglotstop kepmxc\'in, used on analogy with \emph{tk}, as per the footnote on the preceding page, or that this pattern exists as a remnant of an earlier relativization strategy \emph{because} of the influence of N\textbeltl\textschwa\textglotstop kepmxc\'in \emph{tk}. In any case, a syntactic position must be present to house the particle, a position which is also motivated in other languages of the Interior. I presently discuss these other languages.} \subsection{\hspace*{.04in}Evidence from Northern Interior Salish: N\textbeltl\textschwa\textglotstop kepmxc\'in\\ \hspace*{0in} and Secwepemcts\'in} For N\textbeltl\textschwa\textglotstop kepmxc\'in, spoken to the north and west of Okanagan country, oblique markers nearly always introduce relative clauses.\footnote{Exceptions are as follows: (i) before the remote determiner \emph{\textbeltl} where \emph{t} phonologically reduces \citep{Kroeber:1997,Koch:2006}; (ii) in the case of locative relatives, where a preposition introduces the clause; and (iii) in the case of headless relative clauses \citep[fn5]{Koch:2006}. Interestingly, \emph{t} does surface before headless relatives in Nxa\textglotstop amxcín and Okanagan.}$^,$\footnote{The oblique marker \emph{t} in N\textbeltl\textschwa\textglotstop kepmxc\'in is segmentable from the `specific' and `unrealized' determiners \emph{e} and \emph{k}. In keeping them together, I follow the convention of \citet{ThompsonThompson:1992}, who analyze \emph{t} and \emph{k} as a ``single descriptive marker'' (p.153), for example.} Similar to the facts in Okanagan, N\textbeltl\textschwa\textglotstop kepmxc\'in oblique \emph{t} introduces intransitive objects and ditransitive themes, and not transitive objects in main clause contexts. It does occur before relative clauses, however, even in cases where a transitive object has been extracted (\ref{7b}a). \pex[labeloffset=.13in,textoffset=1.5ex]\label{7b} \a\begingl \gla {(w)\textglotstop\'ex} {xe\textglotstop} {cu-t-$\oslash$-\'ene} {e} {z\'e\g{w}tn} \textbf{{\color{blue}{t}}-e} {m\'a\g{\textrevglotstop}-t-st-$\oslash$-ne}.// \glb \textsc{prog} \textsc{dem} fix-\textsc{tr}-\textsc{3sg.obj-1sg.erg} \textsc{det} cup \textsc{obl-det} break-\textsc{im-caus-3sg.obj-1sg.erg}// \glft I am fixing the cup that I broke. \citep[141]{Koch:2006} // \endgl \a\begingl \gla ... {e} {he\textglotstop use\textglotstop} \textbf{{\color{blue}{t}}-k} xwuy' n-s-\textbeltl a\textglotstop\v{x}ans.// \glb ... \textsc{det} egg \textsc{obl-irr} \textsc{fut} \textsc{1sg.poss-nom}-eat(\textsc{intrans})// \glft (I boiled) an egg that I'm going to eat. (Koch, p.c.)// \endgl \xe Since the relative clause predicates \emph{m\'a\g{\textrevglotstop}tstne} ``I broke x'' in (\ref{7b}a) is formally transitive, the oblique preposition \emph{t} cannot have raised with the determiner \emph{e} from a clause-internal position.\footnote{The present determiner \emph{(h)e} coalesces with the oblique marker \emph{t} after movement, to form \emph{te}. The intransitive predicate \emph{\textbeltl a\textglotstop\v{x}ans} ``eat'' in (\ref{7b}b) will take an oblique marked object in main clause contexts, including optionally the irrealis determiner \emph{k} (Koch, p.c.), and so it is possible to analyze \emph{tk} in (\ref{7b}) as introducing a PP which has raised from a post-clausal position, on analogy with N\textbeltl\textschwa\textglotstop kepmxc\'in (\ref{1}) and Okanagan passive agent and instrument extractions. In other words, \emph{t} in (\ref{7b}b) carries nominal case information, and has coalesced with the \emph{t} which normally introduces relative clauses in N\textbeltl\textschwa\textglotstop kepmxc\'in.} As \citet[133]{Koch:2006} notes, ``there must be some higher position, possibly an adjunct to CP, containing the oblique marker \emph{t}'', which he labels XP. This relevant structure is represented by (\ref{7c}): \ex\label{7c} \scriptsize{\Tree[.DP [.D {e} ] [.NP [.NP [.N {z\'e\g{w}tn$_j$} ] ] [.{\color{blue}{XP}} [.{\color{blue}{X}} \textbf{\color{blue}{t}} ] [.CP [.Spec [.DP$t_i$ [.D {e} ] [.NP [.N {$pro_j$} ] ] ] ] [.C' [.C {\ensuremath{\oslash}} ] [.TP ... [.VP {m\'a\g{\textrevglotstop}-t-st-$\oslash$-ne} [.DP$_i$ ] ] ] ] ] ] ] ]} \xe %HD: I have a couple of problems with XP: (i) It looks like in (26) the t is in a *lower* position than the moved D, not a higher one: and you can't just attribute that to phonological reordering, given the intervening pronoun; (ii) the XP potentially blocks the matching relation between head and the moved NP in [SPEC CP]. The second problem is certainly surmountable, but the first indicates to me that the oblique marker is in a *lower* position, within CP, not outside it. How to get it there is another matter. The categorial identity of XP is not immediately important for our purposes, but the existence of an intermediate projection is important.\footnote{Koch (p.c.) is currently investigating the hypothesis that XP is a focus projection, i.e. FocP.} Similar to N\textbeltl\textschwa\textglotstop kepmxc\'in, headed relative clauses in Secwepemcts\'in (a.k.a. Shuswap) are introduced by the oblique case marker \emph{t\textschwa} \citep[67]{Gardiner:1993}. (\ref{7d}a) shows a subject extraction, and (\ref{7d}b) an object extraction: \pex[labeloffset=.13in,textoffset=1.5ex]\label{7d} \a\begingl \gla {\v{c}-l\g{x}-m-st-\'\textepsilon tn} {\textgamma} {sq\'\textepsilon lmx\textsuperscript{w}} \textbf{\color{blue}{t\textschwa}} {wik-t-x}.// \glb \textsc{cust}-know-\textsc{unsp-caus-1sg.erg} \textsc{det} man \textsc{obl} see-\textsc{tr-2sg.erg}// \glft I know the man you saw. \citep[67, ex. 166]{Gardiner:1993}// \endgl \a\begingl \gla {pnh\'\textepsilon\textglotstop n} {k-wik-t-x-w\textschwa s} {\textgamma} {sq\'\textepsilon lmx\textsuperscript{w}} \textbf{\color{blue}{t\textschwa}} {\g{c}\'u\g{m}qs-n-s} {\textgamma} Mary.// \glb when \textsc{irr}-see-\textsc{tr-2sg.erg-3sg.dep} \textsc{det} man \textsc{obl} kiss-\textsc{tr-3sg.erg} \textsc{det} Mary// \glft When did you see the man that kissed Mary? \citep[162]{Gardiner:1993}// \endgl \xe In both cases, the nominal \emph{sq\'\textepsilon lmx\textsuperscript{w}} ``man'' is underlyingly a direct argument of a transitive relative clause. Assuming that Secwepemcts\'in relatives are derived by movement, the prediction is that the clausal remnant should be introduced by one of the three `direct case' determiners: proximal \emph{\textgamma}, distal \emph{l}, or irrealis \emph{k}.\footnote{See \citet[24]{Gardiner:1993} for a discussion of the Secwepemcts\'in determiner system. It seems unlikely that the oblique marker \emph{t\textschwa} which introduces relatives in Secwepemcts\'in consists of \emph{t} plus a coalesced determiner (which is the case for N\textbeltl\textschwa\textglotstop kepmxc\'in), at least synchronically, given the phonological shape of Secwepemcts\'in determiners.} The fact that this prediction is \emph{not} upheld means one of two things: (i) In contrast to the other two Northern Interior Salish languages, Secwepemcts\'in relative clauses are not formed by movement; or more likely (ii) the moved determiner elides after the oblique marker. The second hypothesis is supported by several points. Firstly, \citet[339]{Kroeber:1999} presents data showing that Secwepemcts\'in locative extractions may involve `preposition fronting', similarly to N\textbeltl\textschwa\textglotstop kepmxc\'in (\ref{1}):\footnote{Kroeber states that for (\ref{7g}), ``unfortunately, it is not clear whether it should be interpreted as a headed relative clause (modifying \emph{{cpt\'uk\textsuperscript{w}} {[n} {sx\textsuperscript{w}\'uynt]}} `hole [in the ice]'): if it is, then it would indicate that preposition fronting, or something like it, occurs in Secwepemcts\'in.'' As an alternative, he explains that \emph{n} ``at'' may code the adverbial relation of \emph{s\'ex\textsuperscript{w}m-\textschwa s} ``he bathed'' to the main clause predicate.} \ex[textoffset=.10in]\label{7g} \begingl \gla {m-w\'ik-t-s} {\g{\textcrlambda}-\textglotstop\'ene} {\textgamma} {s-c-\textglotstop\'al-cn-s} {\textgamma} {cpt\'uk\textsuperscript{w}} {n} {sx\textsuperscript{w}}\'uynt {w-s} {\textbf{n}} {s\'ex\textsuperscript{w}m-\textschwa s} {\textgamma} {tww\'iwt}.// \glb \textsc{unsp}-see-\textsc{tr-3sg.erg} \textsc{obl}-here \textsc{det} \textsc{nom-stat}-freeze-edge-\textsc{3sg.poss} \textsc{det} hole in ice \textsc{prog-3sg.conjct} at bathe-\textsc{3sg.conjct} \textsc{det} youth// \glft There they saw the frozen edges of the hole in the ice where the youth had bathed. (ShL T8.172)// \endgl \xe Secondly, note that headless relative clauses \emph{are} introduced by a proximal \emph{\textgamma} or distal \emph{l} determiner, and not the oblique marker \emph{t\textschwa} \citep{Gardiner:1993}: \ex[textoffset=.10in]\label{7e} \begingl \gla {\v{c}-l\g{x}-m-st-\'\textepsilon tn} \textbf{l} {wik-t-x}.// \glb \textsc{cust}-know-\textsc{unsp-caus-1sg.erg} \textsc{det} see-\textsc{tr-2sg.erg}// \glft I know the one you saw./I know that you saw him. \citep[67]{Gardiner:1993}// \endgl \xe% HD: it should be possible to find cases where the preposition only encodes the spatio-temporal location of the DP in the relative clause, not the main clause DP. Ask Dwight to elicit them for you! The fact that headless relatives are introduced by determiners, and headed relatives by the oblique marker (or locative marker) might receive explanation under the following scenario: As in Okanagan, the particles which introduce a pre-posed or headless relative in Secwepemcts\'in must be consistent with the selectional restrictions of the main clause predicate. In (\ref{7e}) for example, the clause must be introduced by a determiner, since the entire relative clause head + clausal modifier constituent is a main clause transitive object argument.\footnote{To clarify, this does not mean that the main-clause predicate \emph{selects} the determiner in these cases, since the determiner has moved from a clause-internal position, but only that the determiner is \emph{consistent} with the main-clause predicate's selectional restrictions. This same requirement also holds for Okanagan, and explains why pre-posed relatives are marginal or ungrammatical in certain cases.} If we assume that the moved determiner elides after the oblique marker in headed relatives, but that the oblique marker elides before the moved determiner in headless relatives (because of the aforementioned consistency requirement), then we have a straightforward explanation for the data. Formally speaking, all relatives in Secwepemcts\'in are of category XP, as they are in N\textbeltl\textschwa\textglotstop kepmxc\'in, but the X position is null for headless relatives.\footnote{It is also possible that the determiner \emph{l} in (\ref{7e}) does not have its source from within the relative clause, but rather introduces the containing DP, whose head happens to be null. This derivation is somewhat more complex, and results in an intermediate stage ordering of \textsc{det-obl}, which is unattested in the Northern Interior. In any case, a determiner test similar to that used in \citet{Koch:2006} may help to clarify the issue.} My analysis predicts that either the oblique marker \emph{or} determiner may introduce a clausal remnant, but not both. Indeed, the non-co-occurrence of these particles seems to be a general feature of the Secwepemcts\'in grammar, since locative adjuncts, for example, are introduced only by prepositions, and never with a co-occurring determiner (Gardiner, p.c.). Note that under this analysis, there is no principled reason why a headed relative might not be introduced by a determiner, as occurs in Okanagan. Data from \citet{Kuipers:1974} and \citet{Kroeber:1999} suggest that this pattern is indeed possible (\ref{7f}):\footnote{Gardiner (p.c.) suggests that a direct determiner introduces the clause in (\ref{7f}b) because this is an example of a headless relative clause, but it is currently unclear to me why \emph{cnc\'en\textschwa mn} cannot be analyzed as the overt head.} \pex[labeloffset=.13in,textoffset=1.5ex]\label{7f} \a\begingl \gla {y\textglotstop\'ene} {\v{x}\textschwa\v{x}\'e\textglotstop} {t} {qlm\'ux\textsuperscript{w}} {...} \textbf{\textgamma} {\g{\textcrlambda}x\textsuperscript{w}-nt-es} {\textgamma} {\v{x}y\'um} {t} {k\textsuperscript{w}\'uk\textsuperscript{w}p\g{y}}.// \glb that powerful \textsc{att} person {...} \textsc{det} beat-\textsc{tr-3sg.erg} \textsc{det} big \textsc{att} chief// \glft the clever (powerful) Indian ... who had won against the great chief. (ShL T7.85),\citep[301]{Kroeber:1999}// \endgl \a\begingl \gla {\textglotstop e\textcrlambda} {w\textglotstop\'ex} {n\textgamma\'i\textglotstop} {\textgamma} {cnc\'en\textschwa mn} \textbf{\textgamma} {cw\g{c}\'e\g{w}m\textschwa s} {t} {sql\'e\g{w}}.// \glb \textsc{conj} \textsc{aux} \textsc{dem} \textsc{det} Chinese \textsc{det} panning \textsc{obl} gold// \glft And there were Chinese there, who were panning for gold.\\ \citep[103]{Kuipers:1974}// \endgl \xe It may therefore be the case that it is only a strong preference, not a requirement, that the moved determiner (rather than the oblique marker) elide in the case of a headed relative. This means that in Secwepemcts\'in, similar to N\textbeltl\textschwa\textglotstop kepmxc\'in and Okanagan, both particles are underlyingly present, but that a co-occurence restriction in Secwepemcts\'in prevents both from surfacing simultaneously. %A similar co-occurrence restriction is also present in Okanagan and Nxa\textglotstop amxcín, as we will see, but it does not have a categorical effect in these languages, as it does in Secwepemcts\'in. Clearly, more work is needed on relativization in Secwepemcts\'in, but for the present purposes, it is important simply to take note of two facts: (i) that oblique \emph{t} introduces headed relatives, regardless of the grammatical status of the relativized constituent; and (ii) locative extractions are consistent with a movement analysis. In sum, data show that Northern Interior Salish relatives may be introduced by \emph{t}, crucially in cases where \emph{t} does not code the relation of the gap to the clausal remnant. I now turn to the Southern Interior, and examine data from Nxa\textglotstop amxcín. %% main clause PP infor, gardiner 1996, p. 9; and 1993, p. 13. %``also Gardiner:1993 67:, post-nominal order is fixed, and that headless relatives are introduced by the direct case marker`` (determiner?) \emph{l}. % % \citet[177 (5)]{Gardiner:1996} % This construction is unlike the one in St'at'imcets where the head generally % does not take a determiner, and when it does, it takes the non-referential % determiner. Secondly, notice that both the relative clause and factive % interpretations are not available The constuction is only interpreted as a factive, % and does not constitute a relative clause at all. This provides evidence that % Secwepemctsin lacks the head final relative clause. The factive in (22) is the % basic structure to which NP preposing has occured in (21). % % .... sounds like t is a complementizer here..... in Shw. % (and St' ?) % from 1993, p. 234,235, it looks like both determiner(?) 'l' and determiner(?) 'k' function as complementizers... check Kroeber. % see 1993, p. 24, for run-down on determiners... % % cf t % % ``The oblique case markers \emph{t\textschwa} and \emph{t\textschwa k} mark realis and irrealis % aspect respectively. They have a wide range of syntactic functions. s Several % functions are illustrated in (44-47). In (44-45) the oblique determiner marks the % patient of a middle construction:'' % % sounds like k is the determiner, ( "k" to be a morphologically distinct irrealis determiner (Gibson 1973), bottom of page 25) here, and te is separate, evidence for separate syntactic position, just like in N\textbeltl\textschwa\textglotstop kepmxc\'in. Expect relatives with tek as well, but not sure if there are any.... % page 41, gardiner:1993. cf examples 44/45 for a nice contrasting t / tk pair in main clauses: \subsection{\hspace*{.04in}Evidence from Southern Interior Salish: Nxa\textglotstop amxcín} Nxa\textglotstop amxcín (a.k.a. Moses-Columbian), a sister language to Okanagan, is similar to the Northern Interior languages just discussed, in that an oblique marker \emph{t} introduces relative clauses, as shown in (\ref{8c}a). Pre-nominal relatives are also introduced by \emph{t} in Nxa\textglotstop amxcín (\ref{8c}b), as are headless relatives (\ref{8c}c).\footnote{Oblique marking of ergatives in Nxa\textglotstop amxcín is not limited to passive agents \citep[88]{Willett:2003}, but from examples I have been able to find, the sequence \emph{\textglotstop an\'i} \emph{t} found in (\ref{8c}b) occurs only before relative clauses. This suggests that the oblique marker in (\ref{8c}b) is associated with the clausal remnant, and not the moved DP.} Unlike in Secwepemcts\'in, there is no co-occurrence restriction preventing a determiner and oblique marker from both introducing a relative clause in Nxa\textglotstop amxcín (\ref{8c}b). \pex[labeloffset=.13in,textoffset=1.5ex]\label{8c} \a\begingl \gla {n\'u\v{x}\textsuperscript{w}t} {\textglotstop ac\'i} {sq\textschwa\g{l}tm\'ix\textsuperscript{w}} \textbf{{t}} {c-my-st\'u-n}.// \glb go \textsc{det} man \textsc{obl} \textsc{cust}-know-\textsc{caus-1sg.erg}// \glft The man that I knew left. \citep[97, ex. 62]{Willett:2003}// \endgl \a\begingl \gla {w\'ik\textbeltl n} {\textglotstop an\'i} \textbf{{t}} {\textglotstop acm\'ux\textsuperscript{w}t} {sm\textglotstop\'amm}.// \glb see-\textsc{tr-1sg.erg} \textsc{det} \textsc{obl} \textsc{cust}-laugh woman// \glft I saw the woman who laughed. \citep[100, ex. 74]{Willett:2003}// \endgl \a\begingl \gla {\textglotstop acs\'ux\textsuperscript{w}sn} {\textglotstop an\'i} \textbf{t} {k\textbeltl-\g{c}\textschwa musntx\textsuperscript{w}}.// \glb \textsc{cust}-know-\textsc{caus-1sg.erg} \textsc{det} \textsc{obl} kiss-\textsc{dir-3sg.erg} { }// \glft I know the one that you kissed. \citep[101, ex.77]{Willett:2003}// \endgl \xe The `general' article \emph{\textglotstop an\'i} is used to introduce direct arguments of transitive predicates in Nxa\textglotstop amxcín \citep[84]{Willett:2003},\footnote{Or one of three `deictic' determiners, \emph{\textglotstop ax\'a\textglotstop} 'proximal', \emph{\textglotstop ac\'i} 'non-proximal', and \emph{\textglotstop a\textbeltl\'u\textglotstop} 'distal'. Note their phonological resemblance to demonstratives in Okanagan. The Nxa\textglotstop amxcín data support Kroeber's hypothesis concerning the source of DP-internal prepositions in the Southern Interior, if we assume that the Okanagan determiner \emph{i\textglotstop} was once a bi-syllabic demonstrative-like determiner, similar to Nxa\textglotstop amxcín \emph{\textglotstop an\'i}, but later underwent truncation, which did not occur in Nxa\textglotstop amxcín. \emph{\textglotstop an\'i} may in fact be the only true determiner in Nxa\textglotstop amxcín, since data in \citet{Mattina:2006} show that it, unlike the other determiners, cannot function as a predicate.} while the oblique marker \emph{t}, just as in Okanagan, is used to introduce non-direct arguments, for example intransitive objects, ditransitive themes, and ergative arguments \citep[87]{Willett:2003}. The fact that \emph{t} surfaces before a transitive relative clause like (\ref{8c}), where the object has been extracted, is not consistent with an analysis where \emph{t} has moved from a clause internal position. (\ref{8c}) therefore seems directly parallel to Secwepemcts\'in (\ref{7d}a) and N\textbeltl\textschwa\textglotstop kepmxc\'in (\ref{7b}a), and the initial hypothesis is thus that \emph{t} in (\ref{8c}) is the head of an XP in a structure essentially equivalent to (\ref{7c}). Locative relative clause data, however, show that (\ref{7c}) is not a correct representation of relative clause formation in Nxa\textglotstop amxcín. (\ref{8d}) below show determiner-locative marker sequences introducing the clausal remnant, which is characteristic of locative adjunct extractions in Okanagan (cf \ref{4b}), except that an additional \emph{t} occurs on the \emph{inside} of the moved DP.\footnote{The locative morpheme \emph{lci} is segmentally, and semantically, more complex than what I analyze as locative K heads in Okanagan. Moses also has the simpler locative `prepositions' found in Okanagan, for example \emph{l}. I currently lack data showing a locative relevant clause introduced by \emph{\textglotstop an\'i} \emph{l}, although I predict that these should be possible, since \emph{\textglotstop an\'i} \emph{l} introduces nominal adjunct in main-clause `prepositional-phrase'-like contexts. Locative \emph{lci} resembles more closely the Okanagan demonstrative adverbial \emph{il\'i\textglotstop} ``there'', but must be analyzed differently, since demonstratives in Okanagan do not occur internal to DP, but adjoin to the exterior. I label \emph{lci} as a Loc head, rather than a K head, for this reason. It is possible that \emph{lci} is a clause-initial locative adverb, and has not moved with the determiner, which in turn means that \emph{t} does not mark the left-periphery in these cases but is situated further inside the clause. A further structural revision will be necessary if this is true. It nevertheless seems reasonable to tentatively assume that since \emph{lci} codes the relation of `the rock' to the relative clause predicate in (\ref{8d}a), that it may also have undergone movement with the determiner.} \pex[labeloffset=.13in,textoffset=1.5ex]\label{8d} \a\begingl \gla {q\textsuperscript{w}t\'unt} {\textglotstop an\'i} {x\g{\textcrlambda}ut$_2$} [[\textbf{\textglotstop an\'i} \textbf{lci} [${\oslash}_{NP_2}$]$_{DP_1}$] \textbf{\color{blue}{t}} {\g{k}\textbeltl\g{t}\'ucntn} {\textglotstop in\g{n}\g{n}\'ik\g{m}n} \emph{$t_1$}$_{CP}$].// \glb big-\textsc{stat} \textsc{det} rock \textsc{det} in.there { } \textsc{obl} \textsc{pos}-put.down-\textsc{dir-1sg.erg} \textsc{1sg.poss}-knife { }// \glft The rock under which I laid the knife is big. \citep[99, ex.71]{Willett:2003}// \endgl \a\begingl \gla {n\textschwa st} {\textglotstop ax\'a} {\textglotstop an\'i} {p\textschwa\g{n}p\textschwa\g{n}\'aqs$_2$} [[\textbf{\textglotstop an\'i} \textbf{lci} [${\oslash}_{NP_2}$]$_{DP_1}$] \textbf{\color{blue}{t}} {nal\'ixn} {\textglotstop int\textschwa\g{m}t\textschwa\g{m}\'utn} \emph{$t_1$}$_{CP}$].// \glb {heavy-\textsc{stat}} \textsc{dem} \textsc{det} \textit{p\textschwa\g{n}p\textschwa\g{n}\'aqs} \textsc{det} in.there { } \textsc{obl} \textsc{pos}-put.down-\textsc{tr-1sg.erg} \textsc{1sg.poss}-clothing { }// \glft The {p\textschwa\g{n}p\textschwa\g{n}\'aqs} where I put my clothes is heavy. \citep[99, ex.70]{Willett:2003}// \endgl \xe It is important to note that sequences of determiner-locative-oblique are only possible in extraction contexts in Nxa\textglotstop amxcín, which means that \emph{t} cannot have moved with the locative DP. Together with Okanagan (\ref{7a}) above, (\ref{8d}) constitutes strong evidence that oblique-marked relative clauses in the Southern Interior, just as in the Northern Interior, do not necessarily involve movement of an oblique marker. Southern Interior Salish languages are thus similar to their Northern Interior counterparts with respect to relative clause formation, but with one important difference: the oblique marker which introduces the relative clause surfaces \emph{before} the moved constituent in the Northern Interior, but \emph{after} the moved constituent in the Southern Interior. This difference implies a structural distinction with regards to relative clause formation. I now turn to an analysis of this difference. % \a\begingl % \gla {*kn} {wik-\textschwa m} {\textbf{t}} {sq\textschwa ltm\'ix\textsuperscript{w}} {\textbf{i\textglotstop }} {ks-knx\'it-m-s.}// % \glb \textsc{1sg.abs} see-\textsc{mid} \textsc{obl} man \textsc{det} \textsc{fut}-help-\textsc{(dir)-2sg.obj-3sg.erg}// % \glft I saw a man who will help you.// %4324LL,NS0[det, rel] % \endgl % \xe % % and Nxa\textglotstop amxcín to the South, oblique markers introduce relative clauses, yet do not (necessarily) say anything about the relation between the gap and the clausal remnant. (and in these languages, relatives are marked oblique regardless of whether the clausal remnant is inflected for future or not. Why it survived in this context in Okanagan remains a bit of a mystery...) % Evidence from Nxa\textglotstop amxcín: t 'always' introduces the clause (but no aspectual restrictions on occurrence of t) % Evidence from N\textbeltl\textschwa\textglotstop kepmxc\'in: t 'always' introduces the clause (but no aspectual restrictions on occurrence of t) \section{\hspace*{.15in}Analysis} I propose that in both Northern and Southern Interior languages, the left-periphery of a relative clause is defined by XP, rather than CP, as implied by \citet{Koch:2006}. The difference is that in the Northern Interior languages, a clause-internal DP moves to the specifier position of CP (cf \ref{7c}) while in Southern Interior languages, a clause-internal DP moves to the specifier position of XP. Okanagan (\ref{8e}a,b) represents data given earlier as (\ref{6b}a,b). \pex[labeloffset=.13in]\label{8e} \a{\scriptsize{\Tree[.KP [.K {t} ] [.NP [.NP [.N {sq\textschwa ltm\'ix\textsuperscript{w}_j} ] ] [.{\color{blue}{XP}} [.Spec [.DPt_i [.D {(i\textglotstop)} ] [.KP [.K {\ensuremath{\oslash}} ] [.NP [.N {pro_j} ] ] ] ] ] [.{\color{blue}{X'}} [.{\color{blue}{X}} \textbf{\color{blue}{t}} ] [.CP [.C {\ensuremath{\oslash}} ] [.TP ... [.VP {ksknx\'it\textschwa ms.} [.DP_i ] ] ] ] ] ] ] ]}\\ \vspace{5.5pt} ``...a man who will help you.''} \a{\scriptsize{\Tree[.DP [.D {i\textglotstop} ] [.KP [.K {$\oslash$} ] [.NP [.NP [.N {sq\textschwa ltm\'ix\textsuperscript{w}_j} ] ] [.{\color{blue}{XP}} [.Spec [.DPt_i [.D {i\textglotstop} ] [.KP [.K {\ensuremath{\oslash}} ] [.NP [.N {pro_j} ] ] ] ] ] [.{\color{blue}{X'}} [.{\color{blue}{X}} {\color{blue}{({\textbf{t}})}} ] [.CP [.C {\ensuremath{\oslash}} ] [.TP ... [.VP {ksknx\'it\textschwa ms.} [.DP_i ] ] ] ] ] ] ] ] ]}\\ %\vspace{3.5pt} ``...a man who will help you.''} \xe %HD: see my comments above. This doesn't solve the problem of the non-contiguity of the D and the oblique marker. I think they are in *different* positions, with t clause-internal. This higher landing site (Spec XP) derives the correct ordering between the initial moved determiner and the following relative clause oblique marker, which is the pattern characteristic in the Southern Interior. This structure will also correctly derive locative relative clauses in Okanagan, as well as Nxa\textglotstop amxcín (\ref{8g}): \ex[textoffset=.10in]\label{8g} \scriptsize{\Tree[.DP [.D {\textglotstop an\'i} ] [.KP [.K {$\oslash$} ] [.NP [.NP [.N {x\g{\textcrlambda}ut$_j$} ] ] [.{\color{blue}{XP}} [.Spec [.DPt_i [.D {\textglotstop an\'i} ] [.LocP [.Loc {\textbf{lci}} ] [.NP [.N {pro_j} ] ] ] ] ] [.{\color{blue}{X'}} [.{\color{blue}{X}} \textbf{\color{blue}{t}} ] [.CP [.C {\ensuremath{\oslash}} ] [.TP ... [.VP {\g{k}\textbeltl\g{t}\'ucntn{ }\textglotstop in\g{n}\g{n}\'ik\g{m}n} [.DP_i ] ] ] ] ] ] ] ] ]}\\ \vspace{5.5pt} ``...the rock under which I laid the knife.'' \xe This analysis predicts that determiner-locative-oblique sequences should be possible in Okanagan, just as in Nxa\textglotstop amxcín (\ref{8g}). This is so far unattested in Okanagan, and may be due to a co-occurrence restriction on sequences of case-marking locative and/or oblique markers. In other words, a double case-marker filter\footnote{Similar in spirit to the Double Determiner Filter advocated by Davis (2010) for St'\'at'imcets relative clauses. The double case-marker filter presumably applies only to two sequences of oblique markers in Nxa\textglotstop amxcín, while Okanagan has extended the filter to include sequences of any two case markers.} in Okanagan prevents a sequence of two oblique markers in instrumental and passive agent extractions (\ref{8h}), and a locative-oblique marker sequence in locative adjunct extractions.\footnote{Although if my morpho-phonological analysis of (\ref{7a}) is correct, and proclitics attach to the leftmost element in XP (in this case \emph{t}), the prediction is that determiner-locative-proclitic-oblique sequences should be possible, since the double-case-marker filter would not apply in these cases.} %\footnote{Similar to Secwepemcts\'in, except that in this language, the co-occurrence restriction applies to sequences of determiners and case markers, not multiple sequences of case markers, which are probably nevertheless also bad in Secwepemcts\'in.} \pex[labeloffset=.13in,textoffset=1.5ex]\label{8h} \a\begingl \gla {Mike} {wiks} {i\textglotstop } {tk\textbeltl m\'ilx\textsuperscript{w}$_2$} [[{i\textglotstop } {t} [\ensuremath{\oslash}$_{NP_2}$]$_{DP_1}$] {(*{\color{blue}{t}})} {\g{c}\'um\g{q}s-nt-m} \emph{t$_1$}$_{XP}$].// \glb Mike see-\textsc{(dir)-3sg.erg} \textsc{det} woman \textsc{det} \textsc{obl} { } {\textsc{(*obl)}} kiss-\textsc{dir-pass}// \glft Mike saw the woman he was kissed by.// %4818LL1Co-valuation [condition C, pro, object relative] \endgl \a\scriptsize{\Tree[.DP [.D {i\textglotstop} ] [.KP [.K {$\oslash$} ] [.NP [.NP [.N {tk\textbeltl m\'ilx\textsuperscript{w}_j} ] ] [.{\color{blue}{XP}} [.Spec [.DPt_i [.D {i\textglotstop} ] [.KP [.K {t} ] [.NP [.N {pro_j} ] ] ] ] ] [.{\color{blue}{X'}} [.{\color{blue}{X}} {\color{blue}{(*t)}} ] [.CP [.C {\ensuremath{\oslash}} ] [.TP ... [.VP {\g{c}\'u\g{m}qs-nt-m} [.DP_i ] ] ] ] ] ] ] ] ]}\\ \xe This analysis has the advantage of explaining the overwhelming preference in Okanagan for post-nominal relative clauses, as opposed to their pre-posed equivalents. Although pre-posed relatives are in certain cases possible in Okanagan (cf \ref{4e}b), these are generally marked since the particles which introduce the pre-posed clausal remnant must \emph{also} match the selectional restrictions of the main clause predicate.\footnote{See above my hypothesis for why headed relatives in Secwepemcts\'in are introduced by oblique markers, and headless relatives are introduced by determiners.} To illustrate, consider again (\ref{8h}). The relative clause \emph{{i\textglotstop } {t} {\g{c}\'um\g{q}s-nt-m}} `who he was kissed by' cannot be preposed over the head-containing DP \emph{{i\textglotstop } {tk\textbeltl m\'ilx\textsuperscript{w}}} `the woman' because the sequence \emph{i\textglotstop} \emph{t} cannot introduce an object of a sentence-initial main-clause transitive predicate like \emph{wiks} `He saw x'. In other words, it is necessary for the case marker \emph{t} to delete in order for the entire relative clause + head constituent to be construed as the DP object argument of the main predicate,\footnote{The requirement that the particle(s) which introduce a head + clausal modifier (or clausal modifier + head) clearly reflect the relationship between the constituent as a whole and the main clause predicate is also active in Secwepemcts\'in, and explains why headless relatives like (\ref{7e}) are introduced by a determiner rather than an oblique marker.} but then the relation between the gap and the relative clause is obscured. The following structure is therefore ungrammatical: \ex \scriptsize{\Tree[.DP [.Spec [.XP_k [.Spec [.DPt_i [.D {i\textglotstop} ] [.KP [.K *{\textbf{t}} ] [.NP [.N {pro_j} ] ] ] ] ] [.X' [.X \ensuremath{\oslash} ] [.CP [.C $\oslash$ ] [.TP ... [.VP {\g{c}\'u\g{m}qs-nt-m} [.DP_i ] ] ] ] ] ] ] [.D' [.D {i\textglotstop} ] [.KP [.K {\ensuremath{\oslash}} ] [.NP [.NP [.N {tk\textbeltl m\'ilx\textsuperscript{w}_j} ] ] [.XPt_k ] ] ] ] ]}\\ \vspace{5.5pt} ``...the woman he was kissed by.'' \xe The same reasoning explains why pre-posed locative relatives are ungrammatical in Okanagan. For an example like (\ref{4b}a, \ref{9h}), the entire relative clause + head constituent can only be construed as an object argument of the main clause predicate if the constituent is introduced by \emph{i\textglotstop}. This requires that the locative case marker \emph{tl} delete, but then this obscures the relation between the gap and the relative clause, as well as resulting in the loss of valuable deictic information. (\ref{9h}) is therefore ungrammatical, while the post-nominal equivalent is acceptable. \ex\label{9h} \scriptsize{\Tree[.DP [.Spec [.XP_k [.Spec [.DPt_i [.D {i\textglotstop} ] [.KP [.K *{\textbf{tl}} ] [.NP [.N {pro_j} ] ] ] ] ] [.X' [.X \ensuremath{\oslash} ] [.CP [.C $\oslash$ ] [.TP ... [.VP {{ac-ylt-m\'i-st-l\textschwa x}} [.DP_i ] ] ] ] ] ] ] [.D' [.D {i\textglotstop} ] [.KP [.K {\ensuremath{\oslash}} ] [.NP [.NP [.N {sq\textschwa ltm\'ix\textsuperscript{w}_j} ] ] [.XPt_k ] ] ] ] ]}\\ \vspace{5.5pt} ``...the man they were running away from.'' \xe Interestingly, the pre-posed equivalents of Okanagan (\ref{6b}a,b) \emph{are} acceptable: \ex\label{9i} \scriptsize{\Tree[.DP [.Spec [.XP_k [.Spec [.DPt_i [.D {i\textglotstop} ] [.KP [.K {$\oslash$} ] [.NP [.N {pro_j} ] ] ] ] ] [.X' [.X {\textbf{({\color{blue}{t}})}} ] [.CP [.C $\oslash$ ] [.TP ... [.VP {ksknx\'it\textschwa ms} [.DP_i ] ] ] ] ] ] ] [.D' [.D {i\textglotstop} ] [.KP [.K {\ensuremath{\oslash}} ] [.NP [.NP [.N {sq\textschwa ltm\'ix\textsuperscript{w}_j} ] ] [.XPt_k ] ] ] ] ]}\\ \vspace{5.5pt} ``...the man who will help you.'' \xe Because \emph{t} in these cases does not code the relation between the gap and the clausal remnant (i.e. it does not move as a constituent with the clause-internal DP), but only serves to `optionally' introduce the relative clause, it may easily elide in order for the entire clausal remnant + head DP constituent to be construed as an argument of the main predicate. (\ref{9i}) is therefore acceptable. In sum, Okanagan, like N\textbeltl\textschwa\textglotstop kepmxc\'in, Secwepemcts\'in, and Nxa\textglotstop amxcín, uses \emph{t} to introduce relatives, and this use of \emph{t} is independent of \emph{t} as a case marker, which codes the relation between the gap and the clausal remnant.\footnote{As mentioned before, there are aspectual restrictions on where relative \emph{t} can surface in Okanagan, which does not seem to be characteristic of any of the other languages surveyed here. Additionally, there is no clear `matching' effect between nominal and head-introducing particles in any of the other languages, as there seems to be in Okanagan, but the prediction is that \emph{if} Secwepemcts\'in allowed determiners and locative/oblique markers to co-occur, that a matching effect would also be evident.} Like St'\'at'imcets (Davis, 2010), Nxa\textglotstop amxcín allows pre-nominal relatives, like example (\ref{8c}b) above. These can be derived from a canonical post-nominal structure by simply inverting the order of the NP and the adjoined XP modifier, as shown in (\ref{9j}). \ex\label{9j} \scriptsize{\Tree[.DP [.D \textbf{\sout{\textglotstop an\'i}} ] !\qsetw{0.5cm} [.KP [.K {\ensuremath{\oslash}} ] [.NP [.{\color{blue}{XP}} [.Spec [.DPt_i [.D \textbf{\textglotstop an\'i} ] [.KP [.K {\ensuremath{\oslash}} ] [.NP [.N {pro$_j$} ] ] ] ] ] [.{\color{blue}{X'}} [.{\color{blue}{X}} {\color{blue}{t}} ] [.CP [.C $\oslash$ ] [.TP ... [.VP {\textglotstop acm\'ux\textsuperscript{w}t} [.DP$_i$ ] ] ] ] ] ] !\qsetw{6cm} [.NP [.N {(sm\'a\textglotstop mm)}$_j$ ] !\qsetw{0.5cm} ] !\qsetw{0.5cm} ] !\qsetw{4cm} ] !\qsetw{4cm} ] }\\ \vspace{5.5pt} ``...the woman/one who laughed.'' \citep[100, ex. 74]{Willett:2003} \xe This implies that in Nxa\textglotstop amxcín, head-modifier ordering between NP and XP is in free variation. The NP-final variant will result in a sequence of two determiners, one of which will delete as a result of a Double Determiner Filter \citep[22]{Davis:2010a}.\footnote{Davis formulates this as consisting of two parts: \pex[exno=i,interpartskip=0pt,belowexskip=0pt,aboveexskip=0pt] \a \emph{Double Determiner Filter}\newline *[D$_1$...D$_2$] where no lexical head intervenes between D$_1$ and D$_2$ \a \emph{Determiner Deletion}\newline Delete one of two phonologically adjacent determiners. \xe} Finally, this analysis offers an explanation of why in Nxa\textglotstop amxcín, the optional absolutive case-marker \emph{wa} surfaces to the right of an associated absolutive argument in extraction contexts, but surfaces to the left of the absolutive argument in other contexts \citep{Willett:1996,Willett:2003}. \pex[labeloffset=.13in,textoffset=1.5ex]\label{9a} \a\begingl \gla {m\'a\g{\textrevglotstop}\textsuperscript{w}-s} \textbf{wa} \textbf{nlx\textsuperscript{w}\'atk\textsuperscript{w}tn} {sm\textglotstop\'amm}.// \glb break-\textsc{(tr)-3sg.erg} \textsc{abs} pot woman// \glft `The woman broke the pot.' \citep[114, ex. 157]{Willett:2003}// \endgl \a\begingl \gla {c-mi-st\'u-nn} {\textglotstop an\'i} \textbf{tt\g{w}\'it$_2$} [[{\textbf{wa}} [$\oslash$ [${\oslash}_{NP_2}$]$_{DP}$]$_{KP_1}$] {t} {k\textbeltl\g{c}\textschwa m\'us-nt-s} {ki\g{\textrevglotstop}\'ana\textglotstop} {\emph{$t_1$}$_{XP}$]}.// \glb \textsc{cust}-know-\textsc{tr-1sg.erg} \textsc{det} boy \textsc{abs} \textsc{det} { } \textsc{obl} kiss-\textsc{tr-3sg.erg} girl { }// \glft `I know the boy that the girl kissed.' \citep[97, ex. 63]{Willett:2003}// \endgl \xe In Nxa\textglotstop amxcín, determiners are regularly absent before argument nominals \citep{Mattina:2006}, as shown in (\ref{9a}a). The categorical status of the moved constituent in (\ref{9a}b) is thus called into question. As opposed to the Nxa\textglotstop amxcín locative markers, \emph{wa} is not a DP-internal K-head since it may apparently precede a determiner in main clause contexts. It introduces both simple absolutive DPs (\ref{9b}a), as well as headless relative clauses which are absolutive DP arguments of main-clause predicates (\ref{9b}b):\footnote{`His grandmother' in (\ref{9b}a) is the subject argument of an intransitive predicate, hence the absolutive marking on the subject, and the oblique marking on the notional object. For (\ref{9b}b), it is unclear whether \emph{wa} has moved with the clause internal DP (cf \ref{9a}b) or indicates that the entire relative clause is an absolutive argument of the main-clause predicate. If the latter holds true, then the prediction is that a secondary \emph{wa} should be grammatical before the DP \emph{\textglotstop an\'i} \emph{tt\g{w}\'it} `the boy' in (\ref{9a}b).} \pex[labeloffset=.13in,textoffset=1.5ex]\label{9b} \a\begingl \gla {\textglotstop ac-y\'ay-\textschwa x\textsuperscript{w}} \textbf{wa} [{\textglotstop a\textbeltl \'u\textglotstop} {kk\'i\g{y}a\textglotstop-s}$_{DP}$] {t} {\v{x}a\v{x}\'{\textschwa}pa\textglotstop}.// \glb \textsc{asp}-weave-\textsc{asp} \textsc{abs} \textsc{det} grandmother-\textsc{3sg.poss} \textsc{obl} bag// \glft `His grandmother is weaving a bag.' \citep[105, ex.14]{Mattina:2006}.// \endgl \a\begingl \gla {\textglotstop\'ica} {k\textsuperscript{w}a} {\textglotstop a\g{c}\v{x}-s} \textbf{wa} [[{\textglotstop a\textbeltl\'u\textglotstop} [${\oslash}_{NP_2}$]$_{DP_1}$] {t} { } { } { } { } { } {\textglotstop ac-x\textsuperscript{w}\'u\g{w}i} {\emph{$t_1$}$_{XP}$]}.// \glb then { } see-\textsc{(tr)-3sg.erg} \textsc{abs} \textsc{det} { } \textsc{obl} { } { } { } { } { } \textsc{asp}-fly { }// \glft `Then he saw the one flying.' \citep[125, ex.103]{Mattina:2006}.// \endgl \xe As indicated by the bracketing in (\ref{9a}b), the examples in (\ref{9b}) show that \emph{wa} is a DP-external case marker \citep{BittnerHale:1996}, and implies that Nxa\textglotstop amxcín has two separate structural positions reserved for case-marking: one is DP-internal, the other is DP-external. Clearly more work is needed on this interesting problem. % \footnote{\citet{Willett:1996} has a workable analysis for why \emph{wa} occurs to the right of a nominal head in quasi-cleft constructions, and \citet[104]{Willett:2003} discusses data showing that \emph{wa} also occurs to the right of a fronted absolutive in so-called `unmarked fronting' structures. Although I am not directly addressing clefts here, I have several issues with her analysis, and since cleft residues are commonly held to be headless relative clauses, a brief discussion may be informative: % % In Nxa\textglotstop amxcín, quasi-clefts consist of a fronted NP, followed by a subordinating particle \emph{\textbeltl u\textglotstop} or \emph{ci}, followed optionally by \emph{wa} if the fronted NP is an absolutive. \citet{Willett:1996} does not elaborate on what exactly the syntactic position of the subordinating particles \emph{\textbeltl u\textglotstop} and \emph{ci} are, but simply has them following the fronted nominal as part of an initial constituent. Consider that subordinating particles \emph{\textbeltl u\textglotstop} and \emph{ci} are cognate with the Kalispel determiners \emph{\textbeltl u\textglotstop} and \emph{ci} respectively. In Kalispel, but not in Nxa\textglotstop amxcín, these determiners introduce arguments in unmarked syntactic contexts. Note also that \emph{\textbeltl u\textglotstop} in particular is segmentally quite similar to the Moses distal determiner \emph{\textglotstop a\textbeltl\'u\textglotstop}. If \emph{\textbeltl u\textglotstop} occurs in determiner position in quasi-clefts, then \emph{wa} can be analyzed straighforwardly as a K head, and a headless relative clause residue, following the structure I have outlined, may be posited for the quasi-cleft. % \ex[exno=i] % \begingl % \gla {s\g{\textcrlambda}ac\'in\textschwa m} {\textbeltl u\textglotstop} {wa} {({\color{blue}{t}})} {\textglotstop a\g{w}t\'aps} {tt\g{w}\'it}.// % \glb deer \textsc{det(?)} \textsc{abs} \textsc{obl} follow-foot-\textsc{(dir)-3sg.erg} boy// % \glft It was the deer that the boy followed.// % \endgl % \xe % This does not explain why \emph{\textbeltl u\textglotstop} does not introduce nominal arguments in an unmarked, main-clause context. % It does seem possible that \emph{\textbeltl u\textglotstop} surfaces in the empty D position of the moved DP after clefting of the NP head, especially given that determiners in Nxa\textglotstop amxcín are often absent. In other words, \emph{\textbeltl u\textglotstop} in Moses is a determiner just as it is in Kalispel, but its distribution is limited to quasi-clefts now. Under this analysis, the other subordinator \emph{ci} could occur in D position. It is notable that the Okanagan subordinator/adjunct-cleft particle, \emph{ki\textglotstop}, is also limited only to introducing quasi-clefts (i.e. adjunct clefts) \citep{Lyon:2010b}.} % \ex\label{8f} % \scriptsize{\Tree[.DP [.D {\textglotstop an\'i} ] [.KP [.K {$\oslash$} ] [.NP [.NP [.N {tt\g{w}\'it_j} ] ] [.{\color{blue}{XP}} [.Spec [.DPt_i [.D {$\oslash$} ] [.KP [.K {\textbf{wa}} ] [.NP [.N {pro_j} ] ] ] ] ] [.{\color{blue}{X'}} [.{\color{blue}{X}} {\color{blue}{t}} ] [.CP [.C {\ensuremath{\oslash}} ] [.TP ... [.VP {{k\textbeltl\g{c}\textschwa m\'usnts} {ki\g{\textrevglotstop}\'ana\textglotstop}} [.DP_i ] ] ] ] ] ] ] ] ]}\\ % \vspace{5.5pt} % % ``...the boy that the girl kissed.'' \citep[97, ex. 63]{Willett:2003} % \xe By way of summary, if we assume that the clause-internal moved constituent in the Southern Interior languages lands in a higher position than in the Northern Interior languages, then with an otherwise identical structure and a minimal amount of extra syntactic machinery, we are able to successfully account for a wide range of relative clauses in Southern Interior Salish, as well as account for the somewhat aberrant distribution of oblique marking in Okanagan. Although relative clause data from Coeur d'Alene and Kalispel are sparse, what may emerge is an interesting split between the Northern and Southern Interior sub-branches of the family with regards to relative clause formation. % but cf moses clefted transitive obejcts, p. 112, where wa occurs outside the relative. % quasi cleft= swat lhu7 / ci wa.... % cleft = swat wa 7alhu7 ... \section{\hspace*{.15in}The relation between DP-internal prepositions and relative clause\\ \hspace*{.12in} formation in the Southern Interior} Southern Interior Salish languages all share the striking property of having DP-internal locative marking, rather than the prepositions found in Northern Interior languages.\footnote{Prepositions in N\textbeltl\textschwa\textglotstop kepmxc\'in and St'\'at'imcets precede a determiner, while prepositions in Secwepemcts\'in never co-occur with a determiner.} In reconstructed Proto-Salish, prepositions also precede determiners,\footnote{The \textsc{prep-det} ordering holds everywhere in the Salish family except the Southern Interior.} and so the question arises as to what caused Southern Interior languages to change the linear ordering of the determiner and preposition. \citet{Kroeber:1999} suggests that determiners in the Southern Interior were originally DP and PP-adjoined demonstratives, which underwent truncation and an accompanying loss of deictic force, thus becoming the determiners that we know today. In the context of a PP, for example, an adjoined demonstrative became a determiner, and the original PP-internal determiners presumably disappeared.\footnote{Nxa\textglotstop amxcín offers the strongest evidence for Kroeber's hypothesis. Both the determiners and demonstratives in this language have a CVCV shape, and so we can easily infer that perhaps Nxa\textglotstop amxcín determiners resisted the truncation which occurred with Okanagan determiners, for example \citep{Mattina:2006}. Alternatively, Nxa\textglotstop amxcín may have completely lost its original determiners, and borrowed a set of demonstrative adverbs to replace them (Davis, p.c.).} I suggest here two possible alternative accounts of how Southern Interior Salish came to have DP-internal locative and oblique marking. Both accounts rest on my analysis of relative clause formation, which I have outlined in this paper. Under the first account, which is not consistent with \citet{Kroeber:1999}, a change in relative clause formation conditioned a linear inversion between case markers and determiners. Under the second account, which is consistent with \citet{Kroeber:1999}, the historical process outlined in the preceding paragraph conditioned a change in relative clause formation. \subsection{\hspace*{.04in}Analysis 1: Change in Relative Clause formation Conditions Inversion} Consider an earlier stage of Okanagan, where relative clause formation occured exactly as in N\textbeltl\textschwa\textglotstop kepmxc\'in, and the language exhibited DP-external prepositions: \pex[labeloffset=.13in,textoffset=1.5ex]\label{8i}\hspace{-.02in}{\textbf{Stage 1:}} Equivalent to N\textbeltl\textschwa\textglotstop kepmxc\'in \a\begingl% \gla {{i\textglotstop }} {s\textschwa x\textsuperscript{w}\g{m}a\textglotstop\g{m}\'aya\textglotstop m-s$_2$} [\textbf{\color{blue}{t}} [[{\textbf{i\textglotstop }} [\ensuremath{\oslash}$_{NP_2}$]$_{DP_1}$] {knx\'it-t-m} \emph{t$_1$} {\textbf{t}} {\textbf{i\textglotstop }} {t\textschwa tw\'it}$_{CP}$]$_{XP}$]// \glb \textsc{det} teacher-\textsc{3sg.poss} \textsc{obl} \textsc{det} { } help-\textsc{dir-pass} { } \textsc{obl} \textsc{det} boy// \glft ...the teacher who was helped by the boy.// %2725LL1[rel] relativized passive agent + loc pp \endgl \a\begingl \gla {{i\textglotstop }} {s\textschwa x\textsuperscript{w}\g{m}a\textglotstop\g{m}\'aya\textglotstop m-s$_2$} [\textbf{\color{blue}{t}} [[{\textbf{(t)}} [{\textbf{i\textglotstop }} [\ensuremath{\oslash}$_{NP_2}$]$_{DP_1}$]$_{PP_1}$] {knx\'it-t-m} \emph{t$_1$} {\textbf{l}} {\textbf{i\textglotstop }} {s\textschwa n\g{q}\textschwa \g{y}m\'int\textschwa n.}$_{CP}$]$_{XP}$]// \glb \textsc{det} teacher-\textsc{3sg.poss} \textsc{obl} \textsc{(obl)} \textsc{det} { } help-\textsc{dir-pass} { } \textsc{loc} \textsc{det} school// \glft ...the teacher that helped him at school.// %2725LL1[rel] relativized passive agent + loc pp \endgl \xe Example (\ref{8i}a) includes an extracted passive patient, and example (\ref{8i}b) includes an extracted passive agent. (\ref{8i}a) additionally has a clause-internal prepositional phrase which is the oblique-marked clause-internal agent. (\ref{8i}b) has a clause internal locative prepositional phrase. Just as in N\textbeltl\textschwa\textglotstop kepmxc\'in today, Okanagan relatives were at one time always introduced by the oblique marker (head X), and a clause-internal DP (or PP) moved to the specifier position of CP. In (\ref{8i}b), which involves extraction of a passive agent, this resulted in a sequence of two oblique markers, only one of which was presumably realized.\footnote{Note that an overt clause-internal oblique-marked agent is necessary in (\ref{8i}a) to confirm the status of the extracted constituent as a patient, since both patient and agent extractions introduce the clausal remnant by the same surface sequence of particles. It is possible that this ambiguity helped motivate an inversion between the determiner and case markers.} At some point during the development of Southern Interior languages, an unknown event prompted a change in relative clause formation. The moved DP (\ref{8j}a) or PP (\ref{8j}b) now lands in the Specifier position of XP, rather than CP: \pex[labeloffset=.13in,textoffset=1.5ex]\label{8j}\hspace{-.02in}{\textbf{Stage 2:}} Movement to Spec XP instead of Spec CP \a\begingl% \gla {{i\textglotstop }} {s\textschwa x\textsuperscript{w}\g{m}a\textglotstop\g{m}\'aya\textglotstop m-s$_2$} [[{\textbf{\textbf{i\textglotstop }}} [\ensuremath{\oslash}$_{NP_2}$]$_{DP_1}$] {\textbf{\color{blue}{t}}} [{knx\'it-t-m} \emph{t$_1$} {\textbf{t}} {\textbf{i\textglotstop }} {t\textschwa tw\'it}$_{CP}$]$_{XP}$]// \glb \textsc{det} teacher-\textsc{3sg.poss} \textsc{det} { } \textsc{obl} help-\textsc{dir-pass} { } \textsc{obl} \textsc{det} boy// \glft ...the teacher who was helped by the boy.// %2725LL1[rel] relativized passive agent + loc pp \endgl \a\begingl%{Stage 2=Movement to Spec XP instead of Spec CP} \gla {{i\textglotstop }} {s\textschwa x\textsuperscript{w}\g{m}a\textglotstop\g{m}\'aya\textglotstop m-s$_2$} [[{\textbf{t}} [{\textbf{\textbf{i\textglotstop }}} [\ensuremath{\oslash}$_{NP_2}$]$_{DP_1}$]$_{PP_1}$] {\textbf{\color{blue}{t}}} [{knx\'it-t-m} \emph{t$_1$} {\textbf{l}} {\textbf{i\textglotstop }} {s\textschwa n\g{q}\textschwa \g{y}m\'int\textschwa n.}$_{CP}$]$_{XP}$]// \glb \textsc{det} teacher-\textsc{3sg.poss} \textsc{obl} \textsc{det} { } \textsc{obl} help-\textsc{dir-pass} { } \textsc{loc} \textsc{det} school// \glft ...the teacher that helped him at school.// %2725LL1[rel] relativized passive agent + loc pp \endgl \xe As can be seen in (\ref{8j}) this change resulted in a discrepancy between the linear order of determiner and preposition found in non-extraction contexts on the one hand (\textsc{prep-det}), and in extraction contexts on the other hand (\textsc{det-prep}). This discrepancy motivated an inversion of determiner and preposition in non-extraction contexts, on analogy with the ordering found before relative clauses: \pex[labeloffset=.13in,textoffset=1.5ex]\label{8k}\hspace{-.02in}{\textbf{Stage 3:}} Inversion of P to case marker, then loss of relative \emph{t} \a\begingl \gla {{i\textglotstop }} {s\textschwa x\textsuperscript{w}\g{m}a\textglotstop\g{m}\'aya\textglotstop m-s$_2$} [[{\textbf{\textbf{i\textglotstop }}} [\ensuremath{\oslash}$_{NP_2}$]$_{DP_1}$] {\sout{\textbf{\color{blue}{t}}}} [{knx\'it-t-m} \emph{t$_1$} {\textbf{i\textglotstop }} {\textbf{t}} {t\textschwa tw\'it}$_{CP}$]$_{XP}$]// \glb \textsc{det} teacher-\textsc{3sg.poss} \textsc{det} { } \textsc{obl} help-\textsc{dir-pass} { } \textsc{det} \textsc{obl} boy// \glft ...the teacher who was helped by the boy.// %2725LL1[rel] relativized passive agent + loc pp \endgl \a\begingl%{Stage 3=Inversion of P to case marker, then loss of relative \emph{t}} \gla {{i\textglotstop }} {s\textschwa x\textsuperscript{w}\g{m}a\textglotstop\g{m}\'aya\textglotstop m-s$_2$} [[{\textbf{i\textglotstop }} [{\textbf{t}} [\ensuremath{\oslash}$_{NP_2}$]$_{KP_1}$]$_{DP_1}$] {\sout{\textbf{\color{blue}{t}}}} [{knx\'it-(t)-m} \emph{t$_1$} {\textbf{i\textglotstop }} {\textbf{l}} {s\textschwa n\g{q}\textschwa \g{y}m\'int\textschwa n.}$_{CP}$]$_{XP}$]// \glb \textsc{det} teacher-\textsc{3sg.poss} \textsc{det} \textsc{obl} { } {\textsc{obl}} help-\textsc{dir-pass} { } \textsc{det} \textsc{loc} school// \glft ...the teacher that helped him at school.// %2725LL1[rel] relativized passive agent + loc pp \endgl \xe The oblique marker which always precedes relative clauses is then lost in cases of passive-patient extraction (\ref{8k}a), since it is now indistinguishable from a case-marker \emph{t} which has become associated with agent-extraction. In agent-extractions (\ref{8k}b), relative \emph{t} merged with the case-marker. Today, relative \emph{t} only surfaces in Okanagan where it cannot be misconstrued as a case-marker: that is, optionally before future marked relatives. \subsection{\hspace*{.04in}Analysis 2: Loss of original determiners conditioned a change in relative\\ \hspace*{.0in} clause formation} The second alternate analysis is consistent with \citet{Kroeber:1999}. We begin again with an earlier stage of Okanagan essentially equivalent to N\textbeltl\textschwa\textglotstop kepmxc\'in (\ref{8i}). The original determiners were lost as DP and PP-adjoined demonstratives evolved into a new set of determiners (in red type), which occurred external to the original prepositions (Stage 2b): \pex[labeloffset=.13in,textoffset=1.5ex]\label{8l}\hspace{-.02in}{\textbf{Stage 2b:}} Loss of original determiners and evolution of original\\ demonstratives into a new set of determiners \a\begingl% \gla {{i\textglotstop }} {s\textschwa x\textsuperscript{w}\g{m}a\textglotstop\g{m}\'aya\textglotstop m-s$_2$} [\textbf{\color{blue}{t}} [[{\textbf{\color{red}{i\textglotstop} }} {\sout{{i\textglotstop} }} [\ensuremath{\oslash}$_{NP_2}$]$_{DP_1}$] {knx\'it-t-m} \emph{t$_1$} {\textbf{\color{red}{i\textglotstop} }} {\textbf{t}} {\sout{{i\textglotstop} }} {t\textschwa tw\'it}$_{CP}$]$_{XP}$]// \glb \textsc{det} teacher-\textsc{3sg.poss} \textsc{obl} \textsc{det} {\sout{\textsc{det}}} { } help-\textsc{dir-pass} { } \textsc{det} \textsc{obl} {\sout{\textsc{det}}} boy// \glft ...the teacher who was helped by the boy.// %2725LL1[rel] relativized passive agent + loc pp \endgl \a\begingl \gla {{i\textglotstop }} {s\textschwa x\textsuperscript{w}\g{m}a\textglotstop\g{m}\'aya\textglotstop m-s$_2$} [\textbf{\color{blue}{t}} [[{\textbf{\color{red}{i\textglotstop} }} [{\textbf{t}} [{\sout{{i\textglotstop} }} [\ensuremath{\oslash}$_{NP_2}$]]$_{PP_1}$]$_{DP_1}$] {knx\'it-t-m} \emph{t$_1$} {\textbf{\color{red}{i\textglotstop} }} {\textbf{l}} {\sout{{i\textglotstop} }} {s\textschwa n\g{q}\textschwa \g{y}m\'int\textschwa n.}$_{CP}$]$_{XP}$]// \glb \textsc{det} teacher-\textsc{3sg.poss} \textsc{obl} \textsc{det} \textsc{obl} {\sout{\textsc{det}}} { } help-\textsc{dir-pass} { } \textsc{det} \textsc{loc} {\sout{\textsc{det}}} school// \glft ...the teacher that helped him at school.// %2725LL1[rel] relativized passive agent + loc pp \endgl \xe During the final stage 3, movement of a clause-internal DP or PP was shifted to a higher position, in order to level the discrepancy between the linear order of determiner and preposition found in non-extraction contexts (\textsc{det-prep}), and that found in extraction contexts (\textsc{prep-det}). This stage is represented above as (\ref{8k}). \subsection{\hspace*{.04in}Weighing the two analyses} The two analyses sketched above differ in the following way: For Analysis 1, the syntax is driving the morphology, while for Analysis 2, the morphology is driving the syntax. Kroeber's hypothesis concerning the origin of the DP-internal locative markers found in the Southern Interior is grounded in typological fact: cross-linguistically, determiners often have their origins in demonstratives \citep{Greenberg:1978}. Furthermore, I have no answer for what factor could possibly motivate the change in relative clause formation proferred by Analysis 1, if \emph{not} the loss of the original determiners, following Kroeber's hypothesis. Although it is possible that there is a precedent for such a change in other languages, I do not know of any. Analysis 2 seems preferable for these reasons. There is at least one piece of evidence that seems to support Analysis 1 over Analysis 2, however, if for no other reason than it either calls into question Kroeber's hypothesis, or introduces new questions concerning the time depth and ordering of the necessary changes. The Secwepemcts\'in determiner \emph{\textgamma(\textschwa)} is probably cognate with the Okanagan determiner \emph{i\textglotstop} (Davis, p.c.). Secwepemcts\'in and Okanagan belong to different sub-branches of the Salish family, which suggests a considerable time depth, but while Okanagan has DP-internal locative markers, Secwepemcts\'in exhibits the remnant of the proto-Salish prepositional system. Although determiners and prepositions/locative markers do not co-occur sequentially in Secwepemcts\'in (Gardiner, p.c.), it does have an oblique-irrealis `article', \emph{t\textschwa k}, which is used to mark non-specific nominals \citep[26]{Gardiner:1993}: \pex[labeloffset=.13in,textoffset=1.5ex] \a\begingl \gla {\g{k}\'ul-m} {t\textschwa} {m\textschwa x\'\textepsilon xy\textschwa\textglotstop}.// \glb make-\textsc{mid} \textsc{obl} basket// \glft She made a basket.// \endgl \a\begingl \gla {m\textepsilon\textglotstop} {\g{k}\'ul-m} {\textschwa k\textsuperscript{w}\textschwa} {t\textschwa k} {m\textschwa x\'\textepsilon xy\textschwa\textglotstop}.// \glb \textsc{exp} make-\textsc{mid} \textsc{rep} \textsc{obl.irr} basket// \glft She's going to make a basket.// \citep[26]{Gardiner:1993} \endgl \xe \emph{t\textschwa k} is cognate with the N\textbeltl\textschwa\textglotstop kepmxc\'in oblique-irrealis determiner complex \emph{tk} found before relative clauses (cf \ref{7b}b). Furthermore, \citet{Gibson:1973} analyzes Secwepemcts\'in \emph{t\textschwa k} as consisting of the oblique marker \emph{t\textschwa} plus the irrealis determiner \emph{k}. Diachronically at least, this certainly seems to be the case, which means that Secwepemcts\'in did exhibit the proto-Salish \textsc{prep-det} ordering at one point during its history, thus placing it in line with the other Northern Interior languages. Assuming cognacy between Okanagan \emph{i\textglotstop} and Secwepemcts\'in proximal determiner \emph{\textgamma\textschwa}, and that \emph{\textgamma\textschwa} occurs in the same syntactic position as the irrealis determiner \emph{k}, we might infer that a proximal oblique in Secwepemcts\'in was once introduced by \emph{*{t\textschwa} {\textgamma\textschwa}}, and a locative oblique by \emph{*{n\textschwa} {\textgamma\textschwa}}, for example. But if both Okanagan and Secwepemcts\'in determiners have evolved from DP/PP-adjoined demonstratives, then we must conclude that for Secwepemcts\'in, an additional set of \emph{locative particles} evolved outside of the new DP domain, and the original locative markers were lost along with the original determiners. This would effectively mean that Secwepemcts\'in was once like Okanagan is today, with DP-internal locative marking, but there is no evidence that Northern Interior languages ever had such structures. %HD: given that in Secwepemcts\'in schwa and e are in complementary distribution (e is simply the stressed form of schwa), it is entirely plausible that Secwepemcts\'in t+ schwa originally looked like N\textbeltl\textschwa\textglotstop kepmxc\'in te, derived from t+ (h)e. So why couldn't OK have lost all its determiners, then borrowed one back from Secwepemcts\'in? JL: Well, it could have, see analysis 2! (I don't say anything about where the determiners come from.) HD: my feeling is that the SI lost its whole D system, then adpoted opportunistically from around it. Secwepemcts\'in is adjacent to OK territory, so provided a source for 7i. And so depending on the strength of the cognate relation between Okanagan \emph{i\textglotstop} and Secwepemcts\'in \emph{\textgamma\textschwa}, Analysis 1 may actually garner some support. \section{\hspace*{.15in}Further Questions} There are many questions which remain, but I will endeavor to address a few of the most salient here. First of all, if Kroeber's hypothesis is incorrect, then we still have an explanation for why Southern Interior Salish languages have DP-internal locative marking (Analysis 1), but suddenly have no explanation for what may have conditioned a change in the way relative clauses are formed. I do not have anything illuminating to say on this issue at the moment, but it may be an avenue worth exploring. Second, what is the categorical status of ``XP'' in Salish, and what syntactic generalizations could follow from assigning X one category label in lieu of another? For the Southern Interior at least, it is possible to analyze X as C; in other words, the oblique marker which introduces relative clauses may be a kind of complementizer. I refrain from making this claim because for Okanagan at least, in contexts involving non-relative clausal subordination, \emph{t} is not found; i.e. it is not used as a complementizer. It is possible that it is assigning oblique status to the CP as a whole, but it is unclear what new generalizations may emerge from this analysis. It is also possible that its categorical status differs within the Southern Interior, since in Okanagan but apparently not in Nxa\textglotstop amxcín, relative \emph{t} is only grammatical before predicates inflected with irrealis \emph{ks-}. I have noted in passing that this could be happening on analogy with N\textbeltl\textschwa\textglotstop kepmxc\'in \emph{tk}, but there is in any case ample historical evidence to posit a separate syntactic position for \emph{t}. Thirdly, the nature of the `matching' relation between the particles which introduce the relative clause head, and the particles which introduce the clausal remnant, remains obscure for Okanagan. Secwepemcts\'in relatives also display evidence for such a matching relation, although the relation is partially obscured by the absence of co-occurring determiners and oblique markers. Finally, extraction data is glaringly scarce on Coeur d'Alene and Kalispel, which makes it difficult to say for sure whether my analysis of Okanagan and Nxa\textglotstop amxcín relatives can truly be extended to the Southern Interior as a whole. But it is nevertheless suggestive that both languages have DP-internal locative markers: \pex[labeloffset=.13in,textoffset=1.5ex]\label{7} {Coeur d'Alene} \citep[ex. 373b, 375]{Doak:1997} \a\begingl \gla {\v{c}n} {d\textepsilon x\textsuperscript{w}-t} {x\textsuperscript{w}\textepsilon} {t\textepsilon\g{l}} {t\g{p}u\g{y}\g{p}u\g{y}\v{s}n}.// \glb \textsc{1sg.nom} drop-\textsc{res} \textsc{det} \textsc{from} car// \glft I fell out of the car.// \endgl \a\begingl \gla {x\textsuperscript{w}\textepsilon} {hnk\textschwa sin} {na\g{q}\textsuperscript{w}c} {x\textsuperscript{w}\textepsilon} {s\g{t}\v{s}\'a} {x\textsuperscript{w}\textepsilon} {t\textepsilon\g{l}} Annie {\v{c}r\textepsilon mqn}.// \glb \textsc{det} \textsc{1sg.poss}-cousin steal-\textsc{dir-3sg.erg} \textsc{det} huckleberry \textsc{det} \textsc{from} Annie Cheremkin// \glft My cousin stole the berries from Annie Cheremkin.// \endgl \xe \pex~[labeloffset=.13in,textoffset=1.5ex]\label{7z} {Spokane} \citep[55]{Carlson:1972}; {Kalispel} \citep[62]{Kroeber:1999} \a\begingl \gla {k\textsuperscript{w}\'eys} {\textbeltl u\textglotstop} {\v{x}\textsuperscript{w}\textschwa\g{l}} {\g{t}a\texthalflength p\textschwa m\'is}// \glb take-\textsc{(dir)-3sg.erg} \textsc{det} \textsc{loc} {arrow-\textsc{3sg.poss}}// \glft He took it for his arrow.// \endgl \a\begingl \gla {k\textsuperscript{w}\'e\g{m}t} {sx\textsuperscript{w}\g{\textcrlambda}\'e\textglotstop i} {\textbeltl u\textglotstop} {l} {\g{t}\'eye\textglotstop} {es\textschwa m\'o\g{q}\textsuperscript{w}}.// \glb {then} {mountain.goat} {\textsc{det}} {\textsc{loc}} {bad} {mountain}// \glft (Then) the mountain goat is in the bad mountain (KL T9.44)// \endgl \xe Kalispel, at least, also shows evidence that locative relatives are formed by movement: \ex[textoffset=.10in] \begingl \gla {x\textsuperscript{w}uy} {\textbeltl a\g{q}\textsuperscript{w}-m-\'ule\textglotstop x\textsuperscript{w}} {\textbeltl u\textglotstop} {l} {es-tix\textsuperscript{w}-i} {t} {s-\v{x}\textsuperscript{w}e\textglotstop\g{l}\'i}.// \glb go come.into.open.space \textsc{det} \textsc{loc} \textsc{cont}-obtain-\textsc{cont} \textsc{obl} camas.in.ground// \glft He came to an open field where people were gathering camas.\\ \citep{Vogt:1940a}, \citep[28]{Camp:2007}// \endgl \xe Further work is needed on relativization in these languages before anything conclusive can be said with regards to the role of oblique marking in relative clauses. \section{\hspace*{.15in}Conclusions} I have claimed that Okanagan and Nxa\textglotstop amxcín show evidence that their relative clauses are formed by movement of a clause-internal DP to the left-periphery of the relative clause. Southern Interior languages thus form their relative clauses in a manner analagous to Northern Interior languages, but not identically. Otherwise anomalous oblique marked relative clauses in Okanagan, and a more general pattern of relativization in Nxa\textglotstop amxcín, together show that the moved constituent lands in a structurally higher position in these languages, than in the Northern Interior languages. Besides representing an interesting typological and syntactic split between two branches of the Salish family, I have suggested that this difference might also explain the DP-internal locative markers characteristic of Southern Interior languages. % % \section*{Appendix: DP-internal `prepositions' in the Southern Interior} % % For Southern Interior Salish languages, particles which correspond to the prepositions found in Northern Interior languages occur internal to the determiner phrase % % The oblique marker likewise occurs internal to DP. % Examples (\ref{7}-\ref{7c}) show that locative markers occur internal to the DP in Coeur d'Alene (\ref{7}), Spokane and Kalispel (\ref{7a}), Nxa\textglotstop amxcín (\ref{7b}), and Okanagan (\ref{7c}). % % \pex\label{7} \textbf{Coeur d'Alene} \citep[ex. 373b, 375]{Doak:1997} % \a\begingl % \gla {\v{c}n} {d\textepsilon x\textsuperscript{w}-t} {x\textsuperscript{w}\textepsilon} {t\textepsilon\g{l}} {t\g{p}u\g{y}\g{p}u\g{y}\v{s}n}.// % \glb \textsc{1sg.nom} drop-\textsc{res} \textsc{det} \textsc{from} car// % \glft I fell out of the car.// % \endgl % \a\begingl % \gla {x\textsuperscript{w}\textepsilon} {hnk\textschwa sin} {na\g{q}\textsuperscript{w}c} {x\textsuperscript{w}\textepsilon} {s\g{t}\v{s}\'a} {x\textsuperscript{w}\textepsilon} {t\textepsilon\g{l}} Annie {\v{c}r\textepsilon mqn}.// % \glb \textsc{det} \textsc{1sg.poss}-cousin steal-\textsc{dir-3sg.erg} \textsc{det} huckleberry \textsc{det} \textsc{from} Annie Cheremkin// % \glft My cousin stole the berries from Annie Cheremkin.// % \endgl % \xe % \pex~\label{7z} \textbf{Spokane} \citep[55]{Carlson:1972}; \textbf{Kalispel} \citep[62]{Kroeber:1999} % \a\begingl % \gla {k\textsuperscript{w}\'eys} {\textbeltl u\textglotstop} {\v{x}\textsuperscript{w}\textschwa\g{l}} {\g{t}a\texthalflength p\textschwa m\'is}// % \glb take-\textsc{(dir)-3sg.erg} \textsc{det} \textsc{loc} {arrow-\textsc{3sg.poss}}// % \glft He took it for his arrow.// % \endgl % \a\begingl % \gla {k\textsuperscript{w}\'e\g{m}t} {sx\textsuperscript{w}\g{\textcrlambda}\'e\textglotstop i} {\textbeltl u} {l} {\g{t}\'eye\textglotstop} {es\textschwa m\'o\g{q}\textsuperscript{w}}.// % \glb {then} {mountain.goat} {\textsc{det}} {\textsc{loc}} {bad} {mountain}// % \glft (Then) the mountain goat is in the bad mountain (KL T9.44)// % \endgl % \xe % \pex~\label{7y} \textbf{Nxa\textglotstop amxcín} \citep[91-92]{Willett:2003},\citep[272-273]{NMattina:2002} % \a\begingl % \gla {\textglotstop acw\'ax-lx} {\g{k}\textschwa\g{m}} {\textglotstop an\'i} {l} {sttx\textsuperscript{w}\'uls} {l} {\textglotstop ana} {sp\textschwa p\'asa\textglotstop sa\textglotstop}.// % \glb \textsc{stat}-live-\textsc{3pl.abs} {still} {\textsc{det}} \textsc{loc} \textsc{nom-ired}-house-\textsc{3sg.poss} \textsc{gen} \textsc{det?} {grandfather}// % \glft They are still living there in that grandfather's little house.// % \endgl % \a\begingl % \gla {kn} {h\textschwa mpmn\v{c}\'ut} {\textglotstop an\'i} {t\g{l}} {in-\textglotstop at\textschwa mup\'il}.// % \glb \textsc{1sg.abs} {get.down.from} \textsc{det} \textsc{loc} \textsc{1sg.poss-car}// % \glft I got out of my car.// % \endgl % \xe % \pex~\label{7x} \textbf{Okanagan} % \a\begingl % \gla {i\textglotstop } {spqm\'ix} {\g{k}\textbeltl k\textsuperscript{w}l'itk\textsuperscript{w}} {i\textglotstop } {l} {\g{t}i\g{k}\textsuperscript{w}t.}// % \glb \textsc{det} swan land.in.water \textsc{det} \textsc{loc} lake// % \glft The swans landed in the lake.// % \endgl % \a\begingl % \gla {m\'iy\textschwa s} {\g{t}iq\textsuperscript{w}\textschwa lq\textsuperscript{w}} {i\textglotstop } {tl} {sc\textschwa c\textglotstop\'ups.}// % \glb more tall \textsc{det} \textsc{loc} {little.sister}-\textsc{3sg.poss}// % \glft She is taller than her little sister.// % \endgl % \xe % % Examples (\ref{8}-\ref{9}) show that the oblique marker also occurs internal to DPs in the Southern Interior. (\ref{8}) shows that in Coeur d'Alene, agents of benefactive applicatives are introduced by the determiner and the oblique marker, and (\ref{8}b) shows that instruments are as well:\footnote{This differs from Upper Nicola Okanagan, which introduces benefactive applicative agents with only the determiner. Colville Okanagan permits benefactive applicative patients (themes) to be introduced by a determiner-oblique marker sequence, as to Coeur d'Alene. \citet[236]{Doak:1997} indicates that benefactive patients are normally introduced by only the oblique marker, which I have found also to be the case for Upper Nicola Okanagan.}$^,$\footnote{The oblique marker \emph{\textglotstop\textepsilon} in Coeur d'Alene is cognate with the prepositional oblique marker \emph{\textglotstop\textschwa} in St'\'at'imcets, rather than the oblique marker \emph{t} found in Okanagan and Nxa\textglotstop amxcín.} (\ref{8a}) shows data from Kalispel and Spokane, and (\ref{8b}), repeated from (\ref{2}), data from Okanagan. Curiously, I could not find any sequences of determiner - oblique marker in Nxa\textglotstop amxcín, \emph{except} in extraction contexts. % % \pex\label{8} \textbf{Coeur d'Alene} \citep[ex. 456, 396]{Doak:1997} % \a\begingl % \gla {\v{c}i\textbeltl\v{s}it\textepsilon lis} {x\textsuperscript{w}\textepsilon} {\textglotstop a} {B\textepsilon rn\'a}// % \glb give-\textsc{appl-1pl.acc-3sg.erg} \textsc{det} \textsc{obl} Bernard// % \glft Barney gave it to us.// % \endgl % \a\begingl % \gla {\v{c}n} {q\textsuperscript{w}\textepsilon lncut} {x\textsuperscript{w}\textepsilon} {\textglotstop\textepsilon} {t\g{p}\'ul\g{p}ul\g{k}\textsuperscript{w}\textepsilon c\textepsilon\textglotstop}.// % \glb \textsc{1sg.nom} burn-\textsc{reflex} \textsc{det} \textsc{obl} cigarette// % \glft I got burned by a cigarette.// % \endgl % \xe % \pex~\label{8a} \textbf{Kalispel} \citep[p.52, 404]{Kroeber:1999}(a,b);\footnote{\citet{Kroeber:1999} glosses \emph{y\'e} and \emph{c\'i} as demonstratives, rather than determiners or articles, possibly because of his historical analysis of these particles. Following \citet{NMattina:2002}, and predecessors, I analyze them as articles. (justify!)} \textbf{Spokane} \citep[88, ]{Carlson:1972,Carlson:1980}(c,d) % \a\begingl % \gla {k\textsuperscript{w}\'e\g{m}t} {w\'e} {x\textsuperscript{w}\'i\gx{\v{c}}-\textschwa\v{s}t-\textschwa m} {y\'e} {t} {s\g{c}\'o\g{m}-s} {y\'e} {sq\'elix\textsuperscript{w}}.// % \glb then in.vain(?) give-\textsc{ditr-ad} \textsc{det} \textsc{obl} bone-\textsc{3sg.poss} \textsc{det} person// % \glft Then they gave him the people's bones to eat. (KL 94 [T9.12])// % \endgl % % \a\begingl kroeber 1999 p. 238 % % \gla {n\'e} {m} {q-s\textschwa-l\'em-t-i} {y\'e} {t} {q-s-ti\v{x}\textsuperscript{w}-s\textschwa nm\'e}.// % % \glb maybe {\textsc{fut}} \textsc{irr-nom}-glad-\textsc{intr-intr.ctn} \textsc{det} \textsc{obl} \textsc{irr-nom}-get.pipe// % % \glft He will be glad to get (this) pipe. (KL T10.38)// % % \endgl % \a\begingl % \gla {u} {c\'i} {t} {\g{p}o\v{x}\textsuperscript{w}\g{p}o\v{x}\textsuperscript{w}\'ut} {u} {c\'i} {t} {\textbeltl-\v{s}e-\v{s}\'u\textglotstop t\textschwa\g{m}} {u} {k\textsuperscript{w}\'e\textglotstop e-i-s} {u} {\textglotstop\'i\textglotstop\textbeltl-i-s}.// % \glb and \textsc{det} \textsc{obl} \textsc{ired}-old and \textsc{det} \textsc{obl} \textsc{dim-ired}-girl and take\textsc{(pl)}-\textsc{dir-3sg.erg} and eat\textsc{(pl)}-\textsc{dir-3sg.erg}// % \glft ...that old man and that girl took it and ate it. (KL T14.71)// % \endgl % \a\begingl % \gla {s\g{p}-nt-\'es} {\textbeltl u\textglotstop} {\v{x}\v{x}\g{\textcrlambda}\textschwa c\'in} {\textbeltl u\textglotstop} {t} {l\'u\g{k}\textsuperscript{w}}.// % \glb hit-\textsc{dir-3sg.erg} \textsc{det} dog \textsc{det} \textsc{obl} stick// % \glft The Indian hit the dog with a stick.// % \endgl % \a\begingl % \gla {x\textsuperscript{w}\'i\g{c}\v{s}t\textschwa n} {\textbeltl u\textglotstop} Agnes {\textbeltl u\textglotstop} {t} {y\'am\v{x}\textsuperscript{w}e\textglotstop}.// % \glb give-\textsc{appl-tr-1sg.erg} \textsc{det} Agnes \textsc{det} \textsc{obl} basket// % \glft I gave a basket to Agnes.// % \endgl % \xe % \pex~\label{9} \textbf{Okanagan} % \a\begingl % \gla {\g{t}\textrevglotstop ap-nt-\'is} [{i\textglotstop } [{t} {swlwlm\'ink}$_{KP}$]$_{DP}$].// % \glb shoot-\textsc{dir-3sg.erg} \textsc{det} \textsc{obl} gun// % \glft He shot it with a gun.// % \endgl % \a\begingl % \gla {Mike} {\g{c}\'u\g{m}qs-nt-m} {i\textglotstop } [{t} {tk\textbeltl m\'ilx\textsuperscript{w}}$_{KP}$]$_{DP}$].// % \glb Mike kiss-\textsc{dir-pass} \textsc{det} \textsc{obl} woman// % \glft Mike was kissed by the woman.// % \endgl % \xe % so moses is the only one not to allow DET - OBL in non-extraction contexts, and (maybe) the only one to allow det - loc - obl sequences before relative clauses.... what does this mean ?? %\pex~\label{8b} \textbf{Nxa\textglotstop amxcín} \citep{} %talk about shuswap ``re'', if it is cognate with i7, and then show that i7 is not a reduced demonstrative..... (possible?) %cf Kinkade 1974 (12) which shows that Moses ``wa'' can co-occur with prepositions! and (20) for possible locative relative with tl'. % Kalispel (Kroeber 1995) marks both passive and non-passive subjects with det-obl. % Coeur d'Alene (Doak 1997) marks ditransitive themes and agents with det-obl. % CONCL- i7 t is not as specialized in the SE as it is in Okanagan. and is VERY specialized in Ms. only occurs in extraction contexts (i.e. obliques are normally just introduced by t, although preps and dets can co-occur). %CHECK try eliciting ``I'm tired of waiting'', in Moses, you get an 'l' preposition % in Cr det obl can mark just about anything, except for an absolutive. non-topic ergative, normal ergative, oblique (Doak 1997, 259-262) Just obl can introduce ergative noun too (510). det -obl can introduce antipassive objects (p. 218,220), instrument, time(?).p.219. %in Cm, only obl introduces ergatives and anti-passive objects, since i7 t is only found in extraction contexts. (that means det either is a main clause det or moved). %in Ka, it looks like det obl can introduce ditransitive themes, normal (non-passive) ergatives, instruments, (See data above). % % \begin{center} % \small{\begin{tabularx}{390pt}{|p{2.7cm}|p{1.22cm}|p{1.3cm}|p{1.35cm}|p{1.65cm}|p{1.13cm}|X|}\hline % & passive agents & other ergative & instrument & benefactive applicative themes & middle objects & locatives\\ \hline % Okanagan\footnote{The determiner is optionally present before passive agents in Okanagan, and I attribute this to phonological reduction. In non-passive contexts in the Upper Nicola dialect, the oblique marker is not normally used, but cf \citet[53]{Kroeber:1999} for data from other dialects, hence \emph{t} is in parentheses. The determiner is not allowed before benefactive applicative themes in the Upper Nicola Dialect, but cf {NMattina:1993} for evidence that the two particles do co-occur in Colville. Anti-passive objects do not allow a determiner in Okanagan, and a determiner reduces before locatives.} & \emph{(i\textglotstop) t} & \emph{i\textglotstop} \emph{(t)}& \emph{(i\textglotstop) t} & \emph{(i\textglotstop) t} & $\oslash$ \emph{(t)} & \emph{(i\textglotstop)} LOC \\ % Nxa\textglotstop amxcín & $\oslash$ \emph{t} & $\oslash$ \emph{t} & ? & $\oslash$ \emph{t} & $\oslash$ \emph{t} & \\ % Coeur d'Alene & \emph{(x\textsuperscript{w}\textschwa){ }\textglotstop\textepsilon} & \emph{(x\textsuperscript{w}\textschwa){ }\textglotstop\textepsilon} & \emph{x\textsuperscript{w}\textschwa{ }\textglotstop\textepsilon} & \emph{x\textsuperscript{w}\textschwa{ }\textglotstop\textepsilon} & \emph{x\textsuperscript{w}\textschwa{ }\textglotstop\textepsilon} & \emph{x\textsuperscript{w}\textschwa{ }\textglotstop\textepsilon} \\ % Spokane-Kalispel & (DET) \emph{t} & DET \emph{t} & DET \emph{t} & DET \emph{t} & (t) & \\ \hline % \end{tabularx}} % \end{center} % \begin{center}\emph{Distribution of Oblique Marking in Southern Interior Salish}\end{center} \subsection*{Abbreviations} \begin{center} \small{\begin{tabular}{|l|l||l|l|}\hline \textsc{abs} & absolutive & \textsc{gen} & genitive object\\ \textsc{appl} & transitive applicative & \textsc{impf} & imperfective\\ \textsc{att} & attributive & \textsc{incept} & inceptive \\ \textsc{aux} & auxiliary & \textsc{indep} & independent pronoun\\ \textsc{caus} & causative transitivizer & \textsc{instr} & instrumental\\ \textsc{cisl} & cislocative & \textsc{intr} & intransitivizer\\ \textsc{comp} & complementizer & \textsc{ired} & initial reduplication\\ \textsc{conj} & conjunction & \textsc{irr} & irrealis\\ \textsc{conjct} & conjunctive & \textsc{loc} & locative\\ \textsc{cont} & continuative & \textsc{mid} & middle marker\\ \textsc{cust} & customary/habitual & \textsc{neg} & negative\\ \textsc{dem} & demonstrative & \textsc{nom} & nominalizer\\ \textsc{deon} & deontic modal & \textsc{obj} & object marker\\ \textsc{dep} & dependent & \textsc{obl} & oblique marker\\ \textsc{det} & determiner & \textsc{pass} & passive\\ \textsc{dir} & directive transitivizer & \textsc{perf} & perfective\\ \textsc{ditr} & ditransitive applicative & \textsc{pl} & plural\\ \textsc{emph} & emphatic & \textsc{pos} & positional\\ \textsc{epis} & epistemic modal & \textsc{poss} & possessive\\ \textsc{erg} & ergative case & \textsc{prog} & progressive\\ \textsc{exis} & assertion-of-existence & \textsc{rep} & reportative\\ \textsc{evid} & evidential & \textsc{sg} & singular\\ \textsc{exp} & expectational & \textsc{stat} & stative\\ \textsc{fred} & final reduplication & \textsc{tr} & transitivizer\\ \textsc{foc} & focus marker & \textsc{unsp} & unspecified\\ \textsc{fut} & future & \textsc{u.poss} & unrealized possessor\\ & & \textsc{ynq} & yes/no question\\ \hline \end{tabular}} \end{center} \vspace{10pt} % \begin{tabularx}{400pt}{|p{2cm}|p{2cm}|p{2cm}|p{2.75cm}|X|} % \hline % {Category} & {Core\newline Argument} & {Adjunct} & {Intrans.Object\newline Ditrans.Patient} & {Predicate}\\ \hline % {DP} & \checkmark & \checkmark & * & *\\ \hline % {KP} & * & \checkmark & \checkmark & {?} \\ \hline % {NP} & * & * & * & \checkmark \\ \hline % \end{tabularx} % Hi Karsten, % % Thanks, those are nice data, I might use one of them in my UVic paper, if that is okay. % % Right, so the main predicate in all those N\textbeltl\textschwa\textglotstop kepmxc\'in examples was intransitive. The head noun is then predictably introduced by "te" (I think), but I believe that your third example I think shows that it isn't just a 'matching' effect between head-introducing and clause-introducing particles. One reason I asked is because Secwepemcts\'in, like N\textbeltl\textschwa\textglotstop kepmxc\'in, also normally introduces relatives with "te". But while N\textbeltl\textschwa\textglotstop kepmxc\'in visible determiner is "he", in Secwepemcts\'in it is "ɣe", so relative "te" in Secwepemcts\'in can't really be argued to consist of an oblique marker + determiner (although maybe diachronically it did). Secwepemcts\'in also has the "tek" before irrealis intransitive objects, (oblique + irrealis det, depending on who you ask) so I'm checking right now with Dwight to see if "tek" occurs before relatives. If so, I think the case can be made that Secwepemcts\'in and N\textbeltl\textschwa\textglotstop kepmxc\'in both share the mysterious "XP" projection you posit in your 2006 paper. That's the point I'm trying to make there. % % In Okanagan, the mysterious "iʔ t" occurs before both relatives which have main clause transitive as well as intransitive predicates. But with a main clause intransitive "t" object, the pattern is (iʔ) t before the relative, while with a main clause transitive "iʔ" object, the pattern is iʔ (t) before the relative. I still don't have an explanation for this matching effect in Okanagan, but my hunch is that it's related to the multple "tk" examples you have in N\textbeltl\textschwa\textglotstop kepmxc\'in. In Okanagan complex DPs (pre-nominal, non-clausal modification), iʔ t never introduces the noun. I guess one prediction is that if Okanagan had a "k" determiner, k t would be possible in complex DP contexts, but this depends on one's analysis of how the prep-det inversion occured. What I was saying at my talk is not consistent with what Kroeber 1999 says, "this peculiarity is readily explained if articles in these languages derive diachronically from demonstrative particles outside DP, or loosely adjoined to it, rather than from articles occupying the determiner slot within DP." I like this explanation, especially because determiners in Moses look exactly like demonstratives, but then what happened to the original determiners inside the PP? I guess I'd like to find some cross-linguistic precedent that a prep-det inversion to det-prep can condition a change in relative clause formation, like I was arguing for. One piece of evidence in my favor, which Kroeber doesn't mention, I think is that Secwepemcts\'in ɣe is cognate with Okanagan iʔ, yet Okanagan iʔ has DP-internal prepositions, while they remain external in Secwepemcts\'in. That may be evidence against an analysis where both ɣe/iʔ developed from reduced demonstratives. % % Probably more detail than you asked for, but that's what is on my mind! % % Thanks for your help, always appreciated. % % John \begin{center}{\textbf{References}}\end{center} \vspace{-35pt} \setlength{\bibsep}{\baselineskip} \renewcommand{\refname}{} {\def\section*#1{} \bibliographystyle{linquiry2}} %make sure path name of bibliography is specified if not in same folder as .tex file try {lsalike} \bibhang=.38in \bibliography{salish} \end{document} % \ex\label{7b} % \begingl % \gla {\textglotstop\'ex} {xe\textglotstop} {\g{q}\textsuperscript{w}\'i\g{y}-es} {he} {s\g{q}\textsuperscript{w}\'iyt} {$\oslash$} {\textbeltl} {x\textsuperscript{w}u\g{y}} {\textglotstop\'upi-$\oslash-\oslash$-s} {\textbeltl} {sk\'ixze\textglotstop-s} {$t_1$}// % \glb \textsc{prog} \textsc{dem} pick-(\textsc{tr})-\textsc{3sg.erg} \textsc{det} fruit \textsc{obl} \textsc{det} \textsc{fut} eat-(\textsc{tr})-\textsc{3sg.erg} \textsc{det} mother-\textsc{3sg.poss} { }// % \glft She is picking the berries that her mother's gonna eat.// % \endgl % \xe % % \ex % \small{\Tree[.DP [.D {he} ] [.NP [.NP [.N {s\g{q}\textsuperscript{w}\'iyt$_j$} ] ] [.{\color{blue}{XP}} [.{\color{blue}{X}} {\color{blue}{t}} ] [.CP [.Spec [.DP$t_i$ [.D {\textbeltl} ] [.NP [.N {$pro_j$} ] ] ] ] [.C' [.C {\ensuremath{\oslash}} ] [.TP ... [.VP {\textglotstop\'upi-$\oslash-\oslash$-s \textbeltl sk\'ixze\textglotstop-s} [.DP$_i$ ] ] ] ] ] ] ] ]} % \xe