<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <p><br>
    </p>
    <p>Hi Johanna, even if I could do this diplomatically, I wouldn't,
      and I think it wouldn't make much sense, as my point is not about
      specific publications or authors; it's about common practice (and
      common practice is reflected in the publications of 'major
      authorities'). But I think I get your point; so let me be a bit
      more specific.</p>
    <p>A lot of (quantitative) typological work relies on 'coding':
      Information is extracted from grammars and transformed into a data
      matrix. Now, it is common practice (and I'm not excluding myself
      here) for the coding to be done by the analyst him/herself, and by
      no one else. But that's considered bad practice in other fields.
      Ideally, you'd need a team of annotators coding independently, on
      the basis of annotation guidelines. The team codes a sample,
      determines inter-annotator agreement, and adjusts/specifies the
      annotation guidelines where necessary. This is done until the
      inter-annotator agreement is satisfactory. And then you can start
      with the actual coding. Ideally, the analyst shouldn't be involved
      in the coding process, as her annotation decisions might be
      (subconsciously) influenced by her working hypotheses. (Note that
      this might be a viable solution to the question of how comparative
      concepts can reliably be defined, for a given study; you can just
      measure how much inter-annotator variation there is; whether or
      not the operationalizations make sense is a different question, of
      course, one of validity. When you use a set of criteria
      disjunctively, the question is what exactly your
      operationalizations are intended to represent.)<br>
    </p>
    <p>Note that I'm not saying that there are no multi-annotator
      projects in typology (I'm actually involved in two such projects,
      though one of them is actually a comparative corpus linguistics
      project); but as far as I can tell, it is 'basically' comon
      practice for analysts to code the data themselves. And I'm not
      saying that mono-annotator projects are useless, sometimes you
      just don't have the manpower for multi-annotator projects (and one
      of the multi-annotator projects I'm involved in was really
      painful; but it was instructive to see that even for categories
      that we thought we had defined rather clearly, inter-annotator
      agreement was rather low in some cases). But as I said earlier, it
      would be nice to have some standards or at least general
      guidelines for coding typological data. Minimally, I think, the
      data should be published, along with at least some information on
      the operational tests that were applied, even if done by a single
      annotator.<br>
    </p>
    <p>I hope this clarifies my (too general) remarks in my previous
      post.<br>
      Volker<br>
    </p>
    <br>
    <div class="moz-cite-prefix">Am 18.11.2017 um 13:27 schrieb Johanna
      NICHOLS:<br>
    </div>
    <blockquote type="cite"
cite="mid:CAHDpjwqRrP9PV8QRwsJJcakZFfzxDHpdHxikP+pCVf6gtd_7Mw@mail.gmail.com">
      <div dir="ltr">
        <div>
          <div>Volker,  <br>
            <br>
          </div>
          If there's a way to do this diplomatically, could you cite an
          example or two of  "important publications by major
          authorities of the field where these criteria are simply not
          applied"?   In linguistics we don't have as much technical
          comment on publications as some other fields do, and maybe we
          should.  In journals where I see technical comments sections
          those comments are refereed, edited, brief, and focused on
          factual and methodological matters, i.e. about empirical
          fundamentals and not debate on theoretical frameworks.</div>
        <div><br>
        </div>
        <div>If there's no way to do it diplomatically, never mind.<br>
        </div>
        <div><br>
        </div>
        Johanna<br>
      </div>
      <div class="gmail_extra"><br>
        <div class="gmail_quote">On Sat, Nov 18, 2017 at 12:37 PM,
          Volker Gast <span dir="ltr"><<a
              href="mailto:volker.gast@uni-jena.de" target="_blank"
              moz-do-not-send="true">volker.gast@uni-jena.de</a>></span>
          wrote:<br>
          <blockquote class="gmail_quote" style="margin:0 0 0
            .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div text="#000000" bgcolor="#FFFFFF">
              <p>Matthew -- are you saying that "one cannot rule out
                disjunctively defined comparative concept" because this
                is what you did?</p>
              <p>I am not convinced by "disjunctive comparative
                concepts". Now, that's nothing for you to worry about --
                I'm just one reader (actually, audience of your
                ALT/2015-talk) who doesn't buy your conclusions because
                he doesn't accept your operationalizations.</p>
              <p>But if we want "to talk TO each other (not only PAST
                each other)", as Martin writes, it would be good to have
                what other fields call "standards of empirical
                research". We have copied a lot of statistical methods
                from fields such as the social sciences and biology. I
                think it would also be beneficial to take a look at
                their standards at the "lower" level -- for instances,
                wrt how data is gathered, processed and classified, how
                hypotheses are operationalized, etc., to make sure that
                the results obtained by somebody are also accepted by
                others (just think of the 5%-threshold for statistical
                significance, which is just a matter of convention).<br>
              </p>
              <p>I'm aware that this type of remark is annoying for some
                of you. I teach both corpus linguistics and typology. In
                corpus linguistics our students deal with very basic
                questions of empirical research -- like the traditional
                'quality criteria' -- e.g. (external, internal)
                validity, objectivity, reliability -- and then, in
                typology, we read important publications by major
                authorities of the field where these criteria are simply
                not applied, sometimes the statistics are faulty, and
                students do enquire about this. What can I say? There
                are no research standards in typology? There is an
                ongoing discussion about "arbitrary/subjective/random/<wbr>disjunctive
                comparative concepts" on the Lingtype-list? I'm afraid
                it wouldn't convince them. What I say is that typology
                still has some way to go to in terms of research
                methods. There are many non-trivial problems, as we have
                seen in various discussions on this list, and we should
                be aware that linguistic data is sui generis (for
                instance, I think we can't adopt just any
                method/software package from genetics). But we shouldn't
                use "authority" as a criterion in our methodological
                choices, and the choices shouldn't be made in such a way
                to legitimize our own research 'ex post'.<br>
              </p>
              <p>Volker<br>
              </p>
              <br>
              <div class="m_-3772461755557999611moz-cite-prefix">Am
                18.11.2017 um 07:36 schrieb Dryer, Matthew:<br>
              </div>
              <blockquote type="cite">
                <div>
                  <p class="MsoNormal">With respect to Martin’s comment</p>
                  <p class="MsoNormal"> </p>
                  <p class="MsoNormal">“<span
                      style="font-size:14.0pt;font-family:Calibri">It is
                      my impression that such ortho-affixes (= forms
                      written as affixes) are perhaps even more common
                      than “phonologically weak” ortho-affixes, but this
                      is an empirical question (in his 2015 ALT
                      abstract, Matthew mentions 248 languages with weak
                      affixes, but 308 languages with only affixes of
                      the Tauya type, apparently confirming my
                      impression).</span>”</p>
                  <p class="MsoNormal"> </p>
                  <p class="MsoNormal">I realize that this is a
                    reasonable inference from my abstract, but one often
                    has to simplify things for the purposes of an
                    abstract. My definition of a weak affix is very
                    narrow and many if not most affixes that are not
                    weak affixes by my narrow criteria can still be
                    shown to be attached phonologically by broader
                    criteria. Furthermore, I also treat a morpheme as an
                    affix for the purposes of this study if it triggers
                    phonologically conditioned allomorphy in stems it
                    attaches to and it is clear from Macdonald’s
                    description of Tauya that some of the ortho-affixes
                    Martin mentions do trigger phonologically
                    conditioned allomorphy in stems they attach to (pp
                    54, 72, 74, 79). </p>
                  <p class="MsoNormal"> </p>
                  <p class="MsoNormal">I counted an affix as weak for
                    the purposes of the study in my 2015 ALT talk only
                    if the description of it in a grammar makes clear
                    that it is nonsyllabic (or has nonsyllabic
                    allomorphs) or that it exhibits phonologically
                    allomorphy or triggers phonologically conditioned
                    allomorphy in adjacent stems. But in many grammars,
                    it is only in the discussion of phonology that it
                    becomes clear that a given affix exhibits
                    phonologically conditioned allomorphy or that it
                    triggers phonologically conditioned allomorphy in
                    adjacent stems. But because I wanted to include a
                    large sample of languages and because it is often
                    unclear from discussions of phonology whether
                    particular rules apply to particular affixes or
                    stems such affixes combine with, I adopted the
                    procedure of not consulting the discussions of
                    phonology in classifying ortho-affixes as weak. This
                    made sense for my 2015 ALT talk since I was
                    examining whether there is a suffixing preference
                    and restricting attention to weak affixes so defined
                    applies equally to prefixes and suffixes. For a
                    different type of typological study, this would have
                    been inappropriate. This illustrates how comparative
                    concepts are specific to particular typological
                    studies.</p>
                  <p class="MsoNormal"> </p>
                  <p class="MsoNormal">Furthermore, there are other
                    factors that I did not examine that are relevant to
                    whether a given ortho-affix is attached
                    phonologically. There may be clear evidence from
                    allophonic rules, but it is often very unclear from
                    grammatical descriptions whether particular
                    allophonic rules apply to particular ortho-affixes
                    or stems to which ortho-affixes are attached. And
                    even if the information is there in the grammatical
                    description, it may take a lot of work to see
                    whether they apply to a particular affix. For
                    example, careful examination of Macdonald’s
                    description of Tauya implies that the benefactive
                    ortho-affix <i>-pe</i> that Martin mentions is
                    attached phonologically, since she gives examples of
                    phonetic representations of forms containing this
                    morpheme where it takes the form [-be] after /m/
                    ([tembe] on page 54).</p>
                  <p class="MsoNormal"> </p>
                  <p class="MsoNormal">There might also be evidence from
                    stress, but still be unclear how stress is assigned
                    to forms including ortho-affixes. For example, Tauya
                    has word-final stress, but it is not clear from
                    Macdonald’s description whether this means that
                    nouns bearing the ortho-affixes that Martin mentions
                    take stress on the ortho-affix.</p>
                  <p class="MsoNormal"> </p>
                  <p class="MsoNormal">Some of you may have noticed that
                    what I say here contradicts what I said in my
                    earlier email about comparative concepts needing to
                    be exhaustive. The comparative concept I used in my
                    2015 ALT talk was not exhaustive and was in fact
                    disjunctive. Since that seemed appropriate for that
                    study, this suggests that one cannot rule out
                    disjunctively defined comparative concepts. I
                    sympathize with Martin’s objecting to disjunctive
                    comparative concepts as a way to continue to use
                    confusing and ambiguous terms and I agree that there
                    is something odd about arbitrary disjunctive
                    comparative concepts, but it is a mistake to simply
                    rule out disjunctive comparative concepts.</p>
                  <p class="MsoNormal"> </p>
                  <p class="MsoNormal">I should note finally that while
                    it is clear that the ortho-affixes that Martin
                    mentions are attached phonologically, they are
                    actually not affixes by either his criteria or mine
                    since they are clitics that attach to postnominal
                    modifiers. [Martin has written about problems with
                    the use of the term “clitic”. I am in complete
                    agreement with him about this. But I use the term
                    here and elsewhere in my research (including my
                    upcoming ALT talk on the encliticization preference)
                    as a label for a comparative concept for grammatical
                    morphemes that are phonologically attached but
                    attach to stems of more than one stem class.]</p>
                  <p class="MsoNormal"> </p>
                  <p class="MsoNormal">Matthew</p>
                </div>
                <div><br>
                </div>
                <span id="m_-3772461755557999611OLK_SRC_BODY_SECTION">
                  <div
style="font-family:Calibri;font-size:11pt;text-align:left;color:black;BORDER-BOTTOM:medium
                    none;BORDER-LEFT:medium
none;PADDING-BOTTOM:0in;PADDING-LEFT:0in;PADDING-RIGHT:0in;BORDER-TOP:#b5c4df
                    1pt solid;BORDER-RIGHT:medium none;PADDING-TOP:3pt">
                    <span style="font-weight:bold">From: </span>Lingtyp
                    <<a
                      href="mailto:lingtyp-bounces@listserv.linguistlist.org"
                      target="_blank" moz-do-not-send="true">lingtyp-bounces@listserv.<wbr>linguistlist.org</a>>
                    on behalf of Martin Haspelmath <<a
                      href="mailto:haspelmath@shh.mpg.de"
                      target="_blank" moz-do-not-send="true">haspelmath@shh.mpg.de</a>><br>
                    <span style="font-weight:bold">Date: </span>Thursday,
                    November 16, 2017 at 7:14 PM<br>
                    <span style="font-weight:bold">To: </span>"<a
                      href="mailto:lingtyp@listserv.linguistlist.org"
                      target="_blank" moz-do-not-send="true">lingtyp@listserv.<wbr>linguistlist.org</a>"
                    <<a
                      href="mailto:lingtyp@listserv.linguistlist.org"
                      target="_blank" moz-do-not-send="true">lingtyp@listserv.<wbr>linguistlist.org</a>><br>
                    <span style="font-weight:bold">Subject: </span>Re:
                    [Lingtyp] wordhood: bonded vs. bound<br>
                  </div>
                  <div><br>
                  </div>
                  <div>
                    <div bgcolor="#FFFFFF" text="#000000"> Matthew Dryer
                      thinks that wordhood is generally understood by
                      grammar authors in terms of <b>bondedness</b> (=
                      phonological weakness, as shown by nonsyllabicity
                      and phono-conditioned allomorphy), not in terms of
                      <b>boundness</b> (= inability to occur in
                      isolation).
                      <p class="MsoNormal"> </p>
                      I don’t know if this is true, but Matthew actually
                      recognizes that grammars often describe
                      grammatical markers as “affixes” even when they do
                      not show the two “phonological weakness” (or
                      bondedness) features.
                      <p class="MsoNormal"> </p>
                      For example, Tauya (a language of New Guinea) is
                      said to have (syllabic) case suffixes, but these
                      never show any allomorphy, e.g.
                      <p class="MsoNormal"> </p>
                      fena’a-ni [woman-ERG]<br>
                      na-pe [you-BEN]<br>
                      wate-’usa [house-INESS]<br>
                      Aresa-nani [Aresa-ALL]<br>
                      Tauya-sami [Tauya-ABL] (MacDonald 1990: 119-126)
                      <p class="MsoNormal"> </p>
                      It is my impression that such ortho-affixes (=
                      forms written as affixes) are perhaps even more
                      common than “phonologically weak” ortho-affixes,
                      but this is an empirical question (in his 2015 ALT
                      abstract, Matthew mentions 248 languages with weak
                      affixes, but 308 languages with only affixes of
                      the Tauya type, apparently confirming my
                      impression).
                      <p class="MsoNormal"> </p>
                      For this reason, I have suggested that the
                      stereotypical “affix” notion should perhaps be
                      captured in terms of boundness together with
                      single-root-class adjacency. Since the Tauya
                      case-markers attach only to nouns, they count as
                      affixes; by contrast, if a bound role marker
                      attaches to both nouns (English “for children”)
                      and adjectives (“for older children”) as well as
                      to other elements (“for many children”), we do not
                      regard it as an affix (but as a preposition), even
                      if it is bound (= does not occur in isolation;
                      English "for" does not).
                      <p class="MsoNormal"> </p>
                      Matthew quite rightly points out that this notion
                      of boundness (which goes back at least to
                      Bloomfield 1933: §10.1) implies that most function
                      words in English are bound, and in fact most
                      function words in most languages are bound – but
                      this is exactly what we want, I feel, because the
                      best way to define a “function word” is as a bound
                      element that is not an affix. Linguists often
                      think of function words (or “functional
                      categories”) as defined semantically, but it is
                      actually very hard to say what is the
                      semantic(-pragmatic) difference between a plural
                      marker and a word like “several”, between a dual
                      marker and the word “two”, between a past-tense
                      marker and the expression “in the past”, or
                      between a comitative marker and the word
                      “accompany”. It seems to me that these
                      distinctions are best characterized in terms of
                      boundness, i.e. inability to occur in isolation.
                      <p class="MsoNormal"> </p>
                      It may be true that occurrence in isolation is a
                      feature of an element that is not easy to elicit
                      from speakers, but in actual language use, there
                      are a very large number of very short utterances,
                      so at least positive evidence for free status
                      (=non-bound status) is not difficult to obtain.
                      <p class="MsoNormal"> </p>
                      In any event, it seems clear to me that some key
                      concepts of grammatical typology such as “flag” (=
                      bound role marker on a nominal) and “person index”
                      (= bound person marker, generally on a verb)
                      require the Bloomfieldian boundness notion, and
                      that these concepts are much easier to work with
                      in typology than the traditional stereotypical
                      notions of “case”, “adposition”, “agreement
                      marker”, and “pronominal clitic”. (For bound
                      person forms, this was a major lesson of Anna
                      Siewierska’s 2004 book “Person”.)
                      <p class="MsoNormal"> </p>
                      Best,<br>
                      Martin<br>
                      <br>
                      <div class="m_-3772461755557999611moz-cite-prefix">On
                        14.11.17 07:02, Dryer, Matthew wrote:<br>
                      </div>
                      <blockquote type="cite">
                        <div>
                          <p class="MsoNormal" style="font-size:14px">I
                            have a number of problems with Martin’s
                            proposal:</p>
                          <p class="MsoNormal" style="font-size:14px"> </p>
                          <p class="MsoNormal" style="font-size:14px"><span
                              style="font-size:14pt">"<b>Here’s a
                                proposal for defining a notion of
                                “affix”, in such a way that the results
                                do not go too much against our
                                intuitions or stereotypes:</b></span></p>
                          <p class="MsoNormal" style="font-size:14px"><span
                              style="font-size:14pt"> </span></p>
                          <p class="MsoNormal" style="font-size:14px"><b><span
                                style="font-size:14pt">An affix is a
                                bound form that always occurs together
                                with a root of the same root-class and
                                is never separated from the root by a
                                free form or a non-affixal bound form."</span></b><span
                              style="font-size:14pt"></span></p>
                          <p class="MsoNormal" style="font-size:14px"> </p>
                          <p class="MsoNormal" style="font-size:14px">If
                            one examines the notion of “bound” from his
                            2013 paper, I believe it implies a
                            comparative concept of affix that differs
                            greatly from what most linguists (at last
                            most non-generative linguists) understand by
                            the term. That’s not a problem for it as a
                            comparative concept, but it is a comparative
                            concept that differs considerably from the
                            stereotype.</p>
                          <p class="MsoNormal" style="font-size:14px"> </p>
                          <p class="MsoNormal" style="font-size:14px">Martin’s
                            definition of “free and “bound” from his
                            2013 paper is as follows:</p>
                          <p class="MsoNormal" style="font-size:14px"> </p>
                          <p class="MsoNormal"><b><span
                                style="font-family:Times;font-size:18px">"But
                                distinguishing in a general way between
                                bound elements and free elements is
                                quite straightforward, because there is
                                a single criterion: Free forms are forms
                                that can occur on their own, i.e. in a
                                complete (possibly elliptical) utterance
                                (Bloomfield 1933: 160). This criterion
                                correlates very highly with the
                                criterion of contrastive use: Only free
                                forms can be used contrastively."</span></b></p>
                          <p class="MsoNormal" style="font-size:14px"><span
                              style="font-size:13.0pt;font-family:Times"> </span></p>
                          <p class="MsoNormal" style="font-size:14px"><span
                              style="font-family:Times">First, I find
                              the notion of complete utterance
                              ambiguous. Does it mean utterances in
                              normal speech or does it include
                              metalinguistic uses (like “What is the
                              last word in the sentence “Who are you
                              going with”? Answer “with”). I would
                              assume that it does not include such
                              metalinguistic uses. But then many if not
                              most so-called function words in English
                              would count as bound since they cannot be
                              used as complete utterances. Perhaps other
                              speakers of English would have different
                              intuitions, but if so that only indicates
                              the lack of clarity in the notion.
                              Furthermore, for many function words in
                              English, I am not sure how to judge
                              whether they can occur alone as
                              utterances. Many such so-called function
                              words would appear to count as bound by
                              Martin’s definition, though they would not
                              count as affixes since they lack other
                              properties in his definition of “affix”.</span></p>
                          <p class="MsoNormal" style="font-size:14px"><span
                              style="font-family:Times"> </span></p>
                          <p class="MsoNormal" style="font-size:14px"><span
                              style="font-family:Times">Second, many
                              languages have grammatical morphemes that
                              must occur adjacent to an open class word
                              but which behave as separate words
                              phonologically. These would all apparently
                              count as affixes by Martin’s definition.
                              Again, I have no problem with this as a
                              comparative concept, only that it means
                              his notion of affix deviates considerably
                              from the stereotype.</span></p>
                          <p class="MsoNormal" style="font-size:14px"><span
                              style="font-family:Times"> </span></p>
                          <p class="MsoNormal" style="font-size:14px"><span
                              style="font-family:Times">Third, Martin
                              says that his criterion “</span><span
                              style="font-family:Times">correlates very
                              highly with the criterion of contrastive
                              use</span><span style="font-family:Times">”.
                              But by my intuitions, the ability to occur
                              as complete utterances does not correlate
                              closely with the criterion of contrastive
                              use, since most so-called function words
                              CAN occur with contrastive use (such as
                              can in this sentence!), as can some
                              morphemes that are conventionally treated
                              as affixes, like <i>un-</i> in “I’m not
                              happy, I’m UNhappy”. Of course, Martin
                              might argue that <i> un-</i> is more like
                              so-called function words and less like
                              morphemes conventionally treated as
                              affixes. But the fact remains that <i>un-</i>
                              is easily the locus of contrast but cannot
                              be used as a complete utterance. I thus
                              see no evidence of a close correlation
                              between the ability to occur as a complete
                              utterance and the ability to be the locus
                              of contrast.</span></p>
                          <p class="MsoNormal" style="font-size:14px"><span
                              style="font-family:Times"> </span></p>
                          <p class="MsoNormal" style="font-size:14px"><span
                              style="font-family:Times">Finally, it is
                              my experience that languages differ in
                              their conventions regarding what can be a
                              complete utterance. Imagine two closely
                              related languages that differ in their
                              grammatical rules governing what is a
                              complete utterance. By Martin’s
                              definition, there might be a large number
                              of morphemes that count as separate words
                              in one language but as affixes in the
                              other language. This strikes me as odd. It
                              seems odd to have a criterion for what is
                              a word and what is an affix so dependent
                              on the grammatical rules in the language
                              for what constitutes a complete utterance.</span></p>
                          <p class="MsoNormal" style="font-size:14px"><span
                              style="font-family:Times"> </span></p>
                          <p class="MsoNormal" style="font-size:14px"><span
                              style="font-family:Times">Matthew</span></p>
                        </div>
                        <div style="font-size:14px"><br>
                        </div>
                        <span
                          id="m_-3772461755557999611OLK_SRC_BODY_SECTION"
                          style="font-size:14px">
                          <div
style="font-family:Calibri;font-size:11pt;text-align:left;color:black;BORDER-BOTTOM:medium
                            none;BORDER-LEFT:medium
none;PADDING-BOTTOM:0in;PADDING-LEFT:0in;PADDING-RIGHT:0in;BORDER-TOP:#b5c4df
                            1pt solid;BORDER-RIGHT:medium
                            none;PADDING-TOP:3pt"> <span
                              style="font-weight:bold">From: </span>Lingtyp
                            <<a
                              href="mailto:lingtyp-bounces@listserv.linguistlist.org"
                              target="_blank" moz-do-not-send="true">lingtyp-bounces@listserv.<wbr>linguistlist.org</a>>
                            on behalf of Martin Haspelmath <<a
                              href="mailto:haspelmath@shh.mpg.de"
                              target="_blank" moz-do-not-send="true">haspelmath@shh.mpg.de</a>><br>
                            <span style="font-weight:bold">Date: </span>Sunday,
                            November 12, 2017 at 10:47 PM<br>
                            <span style="font-weight:bold">To: </span>"<a
href="mailto:lingtyp@listserv.linguistlist.org" target="_blank"
                              moz-do-not-send="true">lingtyp@listserv.<wbr>linguistlist.org</a>"
                            <<a
                              href="mailto:lingtyp@listserv.linguistlist.org"
                              target="_blank" moz-do-not-send="true">lingtyp@listserv.<wbr>linguistlist.org</a>><br>
                            <span style="font-weight:bold">Subject: </span>Re:
                            [Lingtyp] wordhood<br>
                          </div>
                          <div><br>
                          </div>
                          <div>
                            <div bgcolor="#FFFFFF" text="#000000">
                              <p class="MsoNormal">Mattis List and
                                Balthasar Bickel rightly emphasize that
                                “word” is not a Platonic entity (a
                                natural kind) that exists in advance of
                                language learning or linguistic analysis
                                – few linguists would disagree here, not
                                even generativists (who otherwise
                                liberally assume natural-kind
                                catgeories).</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">But I think many
                                linguists still ACT AS IF there were
                                such a natural kind, because the “word”
                                notion is a crucial ingredient to a
                                number of other notions that linguists
                                use routinely – e.g. “gender”, which is
                                typically defined in terms of
                                “agreement” (which is defined in terms
                                of inflectional marking on targets; and
                                inflection is defined in terms of
                                “word”).</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">So is it possible to
                                define a comparative concept ‘word’ that
                                applies to all languages equally, and
                                that accords reasonably with our
                                stereotypes? Note that I didn’t deny
                                this in my 2011 paper, I just said that
                                nobody had come up with a satisfactory
                                definition (that could be used, for
                                instance, in defining “gender” or
                                “polysynthesis”). So I’ll be happy to
                                contribute to a discussion on how to
                                make progress on defining “word”.</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">Larry Hyman notes
                                that other notions like “syllable” and
                                “sentence” are also problematic in that
                                they also “leak”. However, I think it is
                                important to distinguish two situations
                                of “slipperiness”:</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">(1) “Leakage” of
                                definitions due to vague defining
                                notions</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">(2) Incoherence of
                                definitions due to the use of different
                                criteria in different languages</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">The first can be
                                addressed by tightening the defining
                                notions, but the second is fatal.</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">To take up Östen
                                Dahl’s example of the “family” notion:
                                In one culture, a family might be said
                                to be a set of minimally three living
                                people consisting of two adults
                                (regardless of gender) living in a
                                romantic relationship plus all their
                                descendants. In another culture, a
                                family might be defined as a married
                                couple consisting of a man and a woman
                                plus all their living direct ancestors,
                                all their (great) uncles and (great)
                                aunts, and all the descendants of all of
                                these.</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">With two family
                                concepts as different as these, it is
                                obviously not very interesting to ask
                                general cross-cultural questions about
                                “families” (e.g. “How often do all
                                family members have meals together?”).
                                So the use of different criteria for
                                different cultures is fatal here.</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">What I find worrying
                                is that linguists often seem to accept
                                incoherent definitions of comparative
                                concepts (this was emphasized especially
                                in my 2015 paper on defining vs.
                                diagnosing categories). Different
                                diagnostics in different languages would
                                not be fatal if “word” were a Platonic
                                (natural-kind) concept, but if we are
                                not born with a “word” category,
                                typologists need to use the SAME
                                criteria for all languages.</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">So here’s a proposal
                                for defining a notion of “simple
                                morphosyntactic word”:</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal"><b>A simple
                                  morphosyntactic word is a form that
                                  consists of (minimally) a root, plus
                                  any affixes.</b></p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">Here’s a proposal for
                                defining a notion of “affix”, in such a
                                way that the results do not go too much
                                against our intuitions or stereotypes:</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal"><b>An affix is a
                                  bound form that always occurs together
                                  with a root of the same root-class and
                                  is never separated from the root by a
                                  free form or a non-affixal bound form.</b></p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">These definitions
                                make use of the notions of “root” and
                                “root-class” (defined in Haspelmath
                                2012) and<span>  </span>“bound (form)”
                                vs. “free (form)” (defined in Haspelmath
                                2013). All these show leakage as in (1)
                                above, but they are equally applicable
                                to all languages, so they are not
                                incoherent. (I thank Harald Hammarström
                                for a helpful discussion that helped me
                                to come up with the above definitions,
                                which I had not envisaged in 2011.)</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">(What I don’t know at
                                the moment is how to relate “simple
                                morphosyntactic word” to
                                “morphosyntactic word” in general,
                                because I cannot distinguish compounds
                                from phrases comparatively; and I don’t
                                know what to do with “phonological
                                word”.)</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">Crucially, the
                                definitions above make use of a number
                                of basic concepts that apply to ALL
                                languages in the SAME way. David Gil’s
                                proposal, to measure “bond strength” by
                                means of a range of language-particular
                                phenomena, falls short of this
                                requirement (as already hinted by Eitan
                                Grossman). Note that the problem I have
                                with David’s proposal is not that it
                                provides no categorical contrasts
                                (recall my acceptance of vagueness in
                                (1) above), but that there is no way of
                                telling which phenomena should count as
                                measuring bond strength.</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">David’s approach
                                resembles Keenan’s (1976) attempt at
                                defining “subject” (perhaps not by
                                accident, because Ed Keenan was David’s
                                PhD supervisor), but I have a similar
                                objection to Keenan: If different
                                criteria are used for different
                                languages, how do we know that we are
                                measuring the same phenomenon across
                                languages? Measuring X by means of Y
                                makes sense only if we know
                                independently that X and Y are very
                                highly correlated. But do we know this,
                                for subjects, or for bond strength?</p>
                              <p class="MsoNormal"> </p>
                              <p class="MsoNormal">Best,</p>
                              <p class="MsoNormal">Martin</p>
                              <p class="MsoNormal"> </p>
                              <br>
                              <pre class="m_-3772461755557999611moz-signature" cols="72">-- 
Martin Haspelmath (<a class="m_-3772461755557999611moz-txt-link-abbreviated" href="mailto:haspelmath@shh.mpg.de" target="_blank" moz-do-not-send="true">haspelmath@shh.mpg.de</a>)
Max Planck Institute for the Science of Human History
Kahlaische Strasse 10   
D-07745 Jena  
&
Leipzig University 
IPF 141199
Nikolaistrasse 6-10
D-04109 Leipzig    





</pre>
                            </div>
                          </div>
                        </span></blockquote>
                      <br>
                      <pre class="m_-3772461755557999611moz-signature" cols="72">-- 
Martin Haspelmath (<a class="m_-3772461755557999611moz-txt-link-abbreviated" href="mailto:haspelmath@shh.mpg.de" target="_blank" moz-do-not-send="true">haspelmath@shh.mpg.de</a>)
Max Planck Institute for the Science of Human History
Kahlaische Strasse 10   
D-07745 Jena  
&
Leipzig University 
IPF 141199
Nikolaistrasse 6-10
D-04109 Leipzig    





</pre>
                    </div>
                  </div>
                </span> <br>
                <fieldset
                  class="m_-3772461755557999611mimeAttachmentHeader"></fieldset>
                <br>
                <pre>______________________________<wbr>_________________
Lingtyp mailing list
<a class="m_-3772461755557999611moz-txt-link-abbreviated" href="mailto:Lingtyp@listserv.linguistlist.org" target="_blank" moz-do-not-send="true">Lingtyp@listserv.linguistlist.<wbr>org</a>
<a class="m_-3772461755557999611moz-txt-link-freetext" href="http://listserv.linguistlist.org/mailman/listinfo/lingtyp" target="_blank" moz-do-not-send="true">http://listserv.linguistlist.<wbr>org/mailman/listinfo/lingtyp</a><span class="HOEnZb"><font color="#888888">
</font></span></pre>
                <span class="HOEnZb"><font color="#888888"> </font></span></blockquote>
              <span class="HOEnZb"><font color="#888888"> <br>
                  <pre class="m_-3772461755557999611moz-signature" cols="72">-- 
Prof. Volker Gast
English and American Studies
Ernst-Abbe-PLatz 8
D-07743 Jena

Fon: ++49 3641 9-44546
Fax: ++49 3641 9-44542</pre>
                </font></span></div>
            <br>
            ______________________________<wbr>_________________<br>
            Lingtyp mailing list<br>
            <a href="mailto:Lingtyp@listserv.linguistlist.org"
              moz-do-not-send="true">Lingtyp@listserv.linguistlist.<wbr>org</a><br>
            <a
              href="http://listserv.linguistlist.org/mailman/listinfo/lingtyp"
              rel="noreferrer" target="_blank" moz-do-not-send="true">http://listserv.linguistlist.<wbr>org/mailman/listinfo/lingtyp</a><br>
            <br>
          </blockquote>
        </div>
        <br>
      </div>
    </blockquote>
    <br>
    <pre class="moz-signature" cols="72">-- 
Prof. Volker Gast
English and American Studies
Ernst-Abbe-PLatz 8
D-07743 Jena

Fon: ++49 3641 9-44546
Fax: ++49 3641 9-44542</pre>
  </body>
</html>