<!DOCTYPE html>
<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    <p>Many thanks to Jürgen Bohnemeyer for this rich and accurate
      assessment of the issues! I think we don't want to be
      "hard-liners", so it's important to see both advantages and
      downsides of various approaches.</p>
    <p>Clearly, if we were very confident of our deep linguistic
      genealogies (a term I prefer to "phylogenies", because the latter
      term seems to be used in more specialized senses), we should
      definitely use them for causal inference – as we do in biology,
      where we are not only very certain about the higher-level
      groupings (insects, reptiles, birds, mammals, etc), but where more
      recent methods of molecular phylogenetics give us quite
      well-established cladistic trees.</p>
    <p>In linguistics, because of widespread horizontal transfer
      (hybridization, i.e. contact-induced change), the genealogies are
      typically quite uncertain at lower levels – Glottolog has many
      cases of "rake-like" trees (e.g. Southern Bantoid [sout3152] has 9
      branches, and Iranic [iran1269] has eight branches). So do the
      methods adopted from computational biology add all that much? That
      isn't clear to me – maybe also because the papers in the
      general-science journals tend to be short and focused on
      technicalities, not on explaining how the new methods relate to
      earlier methods and why they are better in practice (not only in
      theory).<br>
    </p>
    <p>As Jürgen notes, the correlated-evolution methods (or
      coevolutionary methods, or "phylogenetic methods") are hard to
      carry out in practice, but as Gerhard Jäger observed, it may well
      be that the current relationship between qualitative linguistics
      and quantitative phylogenetics resembles the earlier tension-laden
      relationship between qualitative archeology and Carbon-14 methods
      from physics. So indeed, it could be that linguists will in the
      future rely on a separate set of experts to carry out these
      analyses (if the funding allows it).</p>
    <p>It's good to hear from Simon Greenhill about the problems with
      the sampling method in comparative anthropology – we should
      definitely learn from those experiences. Anthropologists have a
      much longer history of using quantitative methods and thinking
      about them. But I don't understand why the problems with the
      statified sampling method should be worse in linguistics than in
      anthropology – yes, language change proceeds much more slowly, but
      we also have a lot less borrowing in languages. It seems that
      cultural borrowing is all over the place, and it is impossible to
      establish independent anthropological phylogenies.</p>
    <p>Simon mentions a number of theoretical possibilities of sampling
      in linguistics, but those bear little resemblance to what
      typologists have actually been doing and advocating. One of the
      most interesting approaches, in my view, is Matthew Dryer's (2018)
      combination of genealogical and geographical stratification, where
      the geographical distance is not measured in bare space ("200
      miles"), but in terms of languages spoken in between ("30
      languages").</p>
    <p>Best,</p>
    <p>Martin<br>
    </p>
    <p><br>
    </p>
    <div class="moz-cite-prefix">On 06.11.23 00:22, Juergen Bohnemeyer
      wrote:<br>
    </div>
    <blockquote type="cite"
cite="mid:SJ0PR15MB46966D0B2BE32910582F1528DDA4A@SJ0PR15MB4696.namprd15.prod.outlook.com">
      <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
      <meta name="Generator"
        content="Microsoft Word 15 (filtered medium)">
      <style>@font-face
        {font-family:Helvetica;
        panose-1:0 0 0 0 0 0 0 0 0 0;}@font-face
        {font-family:Wingdings;
        panose-1:5 0 0 0 0 0 0 0 0 0;}@font-face
        {font-family:SimSun;
        panose-1:2 1 6 0 3 1 1 1 1 1;}@font-face
        {font-family:"Cambria Math";
        panose-1:2 4 5 3 5 4 6 3 2 4;}@font-face
        {font-family:Calibri;
        panose-1:2 15 5 2 2 2 4 3 2 4;}@font-face
        {font-family:"CMU Serif";
        panose-1:2 0 6 3 0 0 0 0 0 0;}@font-face
        {font-family:"Times New Roman \(Body CS\)";
        panose-1:2 11 6 4 2 2 2 2 2 4;}@font-face
        {font-family:"\@SimSun";
        panose-1:2 1 6 0 3 1 1 1 1 1;}p.MsoNormal, li.MsoNormal, div.MsoNormal
        {margin:0in;
        font-size:10.0pt;
        font-family:"Calibri",sans-serif;}a:link, span.MsoHyperlink
        {mso-style-priority:99;
        color:blue;
        text-decoration:underline;}p.MsoListParagraph, li.MsoListParagraph, div.MsoListParagraph
        {mso-style-priority:34;
        margin-top:0in;
        margin-right:0in;
        margin-bottom:0in;
        margin-left:.5in;
        font-size:10.0pt;
        font-family:"Calibri",sans-serif;}span.EmailStyle19
        {mso-style-type:personal-reply;
        font-family:"CMU Serif";
        color:windowtext;
        font-weight:normal;
        font-style:normal;}.MsoChpDefault
        {mso-style-type:export-only;
        font-size:10.0pt;
        mso-ligatures:none;}div.WordSection1
        {page:WordSection1;}ol
        {margin-bottom:0in;}ul
        {margin-bottom:0in;}</style>
      <div class="WordSection1">
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">Dear
            all – Two points:<o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">First,
            this thread has served to alleviate somewhat my concerns
            about the review processes used by the generic science
            journals – mostly through offline one-on-one communications
            I’ve received. Thanks!<o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">Secondly,
            I’d like to, maybe gently push back, or may be just
            elaborate, on the contrast Simon Greenhill draws between
            phylogenetic inference and the “overly simplistic
            correlation methods”, as he calls them. In the process, I’d
            like to try to clarify what in my mind are the strengths and
            weaknesses of each (type of) method. And I’ll pose a couple
            of challenge questions to the detractors of
            phylogenetic/coevolutionary testing and one for their
            proponents, and I hope all of these questions won’t go
            unanswered </span>
          <span
style="font-size:12.0pt;font-family:"Apple Color Emoji"">😊</span><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">Before
            I do that, let me flag my interest in this: I’m currently in
            Leipzig spending a sabbatical with Russell Gray’s Department
            of Linguistic and Cultural Evolution (</span><span
style="font-size:12.0pt;font-family:"Cambria Math",serif;mso-fareast-language:ZH-CN">≈
          </span><span
            style="font-size:12.0pt;font-family:"CMU Serif"">the
            Grambank team), with the principal objective of figuring out
            how to capitalize the most on Grambank for my project. The
            project tracks the typological distribution of functional
            expressions and attempts to account for it (the
            distribution) in terms of an evolutionary process that makes
            grammaticalization sensitive to functional pressures. Since
            coming here, I’ve been struggling with the question whether
            I should adopt stratified sampling, a phylogenetic method
            (I’m not going to use the term ‘co-evolution’ here, because
            I’m at this stage of the project just looking at the
            emergence of individual functional expressions), or both. So
            I’m airing my thoughts very much in hopes of eliciting
            feedback.</span><span
style="font-size:12.0pt;font-family:"Cambria Math",serif;mso-fareast-language:ZH-CN"><o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">So.
            Let’s start with a very basic reminder: We cannot observe
            causality, we can only infer it, as David Hume taught us
            (i.e., humanity) nearly three centuries ago. The difference
            between the methods at issue is one between synchronic
            inference and diachronic inference.
            <o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">Now,
            before I go any further, a big disclaimer: I think I
            understand at the most abstract level the ideas underlying
            phylogenetic inference. But I don’t understand the
            statistical “sausage making” involved. Meaning I have many
            basic questions about the algorithms used for phylogenetic
            inference.
            <o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">With
            that said: Imagine an ideal world in which we had perfect
            and reliable knowledge about the complete phylogeny of every
            extant human language and every language that ever existed.
            It seems indisputable to me that in this world, we would
            test hypotheses about causal relations between two
            linguistic features not merely by looking at their
            synchronic distributions, but also by looking at whether the
            two features tend to co-evolve across phylogenies or whether
            they appear to have merely accidentally “travelled together”
            through time in a few families.
            <o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">So
            this would be my first question to the detractors of
            phylogenetic/dynamic (Maslova’s term)/co-evolutionary tests
            for typological generalizations: do you disagree that if we
            had that perfect knowledge, we would as a matter of course
            take into account the phylogenetic perspective?<o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">Now,
            in the real world, our knowledge of phylogenies is itself
            largely inferred from synchronic distributions. Crucially,
            though, these inferences are based on data that may overlap
            with the typological patterns of interest, but is largely
            independent of it. To make this more concrete:<o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <ul style="margin-top:0in" type="disc">
          <li class="MsoListParagraph"
            style="margin-left:0in;mso-list:l0 level1 lfo1"><span
              style="font-size:12.0pt;font-family:"CMU Serif"">Jäger
              & Wahle (2021) use the world tree inferred from
              <a href="https://asjp.clld.org/" moz-do-not-send="true">ASJP</a>
              cognate data presented in Jäger (2018). Notably, Jäger’s
              method takes into account systematic sound
              correspondences, so to my uninitiated eye, it looks like
              an unassailable computational implementation of the
              historic-comparative method. Of course, the data the
              analysis is based on is another matter.<o:p></o:p></span></li>
        </ul>
        <p class="MsoListParagraph"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <ul style="margin-top:0in" type="disc">
          <li class="MsoListParagraph"
            style="margin-left:0in;mso-list:l0 level1 lfo1"><span
              style="font-size:12.0pt;font-family:"CMU Serif"">Verkerk
              et al. (ms.), which Martin cites in the post that started
              the thread, rely on Glottolog, which Jäger & Wahle
              also used to validate their inferred world tree. In my
              understanding, Glottolog is sourced by a compilation of
              the best available evidence for phylogenetic relations,
              principally curated by its lead author, Harald
              Hammarström.<o:p></o:p></span></li>
        </ul>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">Two
            more points to keep in mind here:<o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <ul style="margin-top:0in" type="disc">
          <li class="MsoListParagraph"
            style="margin-left:0in;mso-list:l0 level1 lfo1"><span
              style="font-size:12.0pt;font-family:"CMU Serif"">Stratified
              sampling of course also takes into account phylogenetic
              and areal information. So the two types of approaches have
              really more in common than Simon gives them credit for in
              his reply, I think. However, stratified sampling uses
              phylogenetic and areal information in first approximation
              only “negatively”, as it were: by systematically removing
              observations from the analysis that are suspected to be
              “tainted” by areal or genealogical dependencies. <o:p></o:p></span></li>
        </ul>
        <p class="MsoListParagraph"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <ul style="margin-top:0in" type="disc">
          <li class="MsoListParagraph"
            style="margin-left:0in;mso-list:l0 level1 lfo1"><span
              style="font-size:12.0pt;font-family:"CMU Serif"">The
              distance of our real world from that ideal world in which
              we have perfect knowledge of phylogenetic and areal
              relations is in my (again, uninitiated, in the sense that
              I’m not a historical linguist) mind pretty darn great.
              However, the phylogenetic statistics algorithms use
              “forests” of alternative phylogenetic trees weighted for
              the confidence the field (or the analysts) has/have in
              them, factored into the tests as Bayesian priors. In this
              sense, the use of phylogenetic information is more
              sophisticated in the dynamic tests than in anything based
              on stratified sampling that I have seen.<o:p></o:p></span></li>
        </ul>
        <p class="MsoListParagraph"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">So
            with all this assembled, here’s my semi-informed take on the
            pros and cons of each approach:<o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <ul style="margin-top:0in" type="disc">
          <li class="MsoListParagraph"
            style="margin-left:0in;mso-list:l0 level1 lfo1"><span
              style="font-size:12.0pt;font-family:"CMU Serif"">Stratified
              sampling (I’m going to start with the method that has been
              practiced the longest)<o:p></o:p></span></li>
        </ul>
        <p class="MsoListParagraph"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <ul style="margin-top:0in" type="disc">
          <ul style="margin-top:0in" type="circle">
            <li class="MsoListParagraph"
              style="margin-left:0in;mso-list:l0 level2 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Pro<o:p></o:p></span></li>
            <ul style="margin-top:0in" type="square">
              <li class="MsoListParagraph"
                style="margin-left:0in;mso-list:l0 level3 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Technologically
                  and conceptually simple, easy to implement by anybody
                  with the most basic level of training in inferential
                  statistics, and transparent for the analyst to track
                  and understand the effects of their decisions.<o:p></o:p></span></li>
              <li class="MsoListParagraph"
                style="margin-left:0in;mso-list:l0 level3 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Statistically
                  conservative, in the specific sense that it minimizes
                  the odds of false positives, i.e., of support for
                  generalizations for which the evidence isn’t really
                  there. (One might think that Dunn et al. 2011 present
                  counterevidence to this. But the apparent false
                  positives pointed out by Dunn et al. turn out to be
                  true positives after all, now that Jäger & Wahle
                  and Verkerk et al. have replicated their analysis with
                  a larger sample.)<o:p></o:p></span></li>
              <li class="MsoListParagraph"
                style="margin-left:0in;mso-list:l0 level3 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Validity of
                  testing depends only on one key assumption – the big
                  one: independence of observations. (Detractors of
                  stratified sampling might however argue that it is
                  impossible to ensure that this assumption is warranted
                  in any decent-sized sample.)<o:p></o:p></span></li>
            </ul>
            <li class="MsoListParagraph"
              style="margin-left:0in;mso-list:l0 level2 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Con<o:p></o:p></span></li>
            <ul style="margin-top:0in" type="square">
              <li class="MsoListParagraph"
                style="margin-left:0in;mso-list:l0 level3 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">There has
                  never been a consensus on an optimal sampling
                  algorithm that balances minimization of areal and
                  genealogical biases with retaining enough statistical
                  power to detect the patterns of interest. With such a
                  consensus absent after 45 years of some of the
                  smartest minds in the field applying themselves to the
                  problem, maybe there is no optimal solution to be had?<o:p></o:p></span></li>
              <li class="MsoListParagraph"
                style="margin-left:0in;mso-list:l0 level3 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Based on
                  information reduction (exclusion of observations);
                  does not take into account phylogenetic information to
                  the fullest measure of what can be concluded from it.<o:p></o:p></span></li>
            </ul>
          </ul>
        </ul>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <ul style="margin-top:0in" type="disc">
          <li class="MsoListParagraph"
            style="margin-left:0in;mso-list:l0 level1 lfo1"><span
              style="font-size:12.0pt;font-family:"CMU Serif"">Dynamic/phylogenetic/co-evolutionary
              inference<o:p></o:p></span></li>
        </ul>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <ul style="margin-top:0in" type="disc">
          <ul style="margin-top:0in" type="circle">
            <li class="MsoListParagraph"
              style="margin-left:0in;mso-list:l0 level2 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Pro<o:p></o:p></span></li>
            <ul style="margin-top:0in" type="square">
              <li class="MsoListParagraph"
                style="margin-left:0in;mso-list:l0 level3 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Utilizes the
                  best available phylogenetic information “for all it’s
                  worth”, directly examining whether features tend to
                  coevolve across lineages or only in particular
                  lineages. (The same cannot currently be said about
                  areal information. However, Verkerk et al. (ms.) use
                  mixed models regression – which is feasible based on
                  Grambank thanks to its relatively low percentage of
                  missing observations – to factor in possible areal
                  effects.)<o:p></o:p></span></li>
              <li class="MsoListParagraph"
                style="margin-left:0in;mso-list:l0 level3 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Does not
                  depend on the assumption of independence of
                  observation.<o:p></o:p></span></li>
            </ul>
          </ul>
        </ul>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <ul style="margin-top:0in" type="disc">
          <ul style="margin-top:0in" type="circle">
            <li class="MsoListParagraph"
              style="margin-left:0in;mso-list:l0 level2 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Con<o:p></o:p></span></li>
            <ul style="margin-top:0in" type="square">
              <li class="MsoListParagraph"
                style="margin-left:0in;mso-list:l0 level3 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Very steep
                  training demands both on the conceptual and on the
                  technological side. Currently only a handful of
                  specialists are able to perform such analyses. In
                  addition, the computational complexities are so
                  demanding as to render the analyses effectively opaque
                  for all but those specialists, raising questions about
                  the validation of such tests. (Gerhard Jäger’s
                  comparison to carbon dating in archeology is apt here
                  as well, I think: while phylogenetic analysis will
                  likely become more standard in the future, it may be
                  the case that typologists will always have to rely on
                  specialists for actually performing such tests, just
                  like archeologists rely on physicists or lab
                  technicians for performing carbon dating. I think </span>
                <span
style="font-size:12.0pt;font-family:"Apple Color Emoji"">😉</span><span
style="font-size:12.0pt;font-family:"CMU Serif"">)<o:p></o:p></span></li>
              <li class="MsoListParagraph"
                style="margin-left:0in;mso-list:l0 level3 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Reliability
                  of phylogenetic inference seems to depend on the size
                  of the phylogenies involved: the smaller the family,
                  the higher the uncertainty in the absence of actual
                  historic data.<o:p></o:p></span></li>
              <li class="MsoListParagraph"
                style="margin-left:0in;mso-list:l0 level3 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">As with
                  stratified sampling, there does not yet appear to be a
                  consensus approach.
                </span><span
style="font-size:12.0pt;font-family:"CMU Serif"" lang="DE">Jäger
                  & Wahle (2021) and Verkerk et al.
                </span><span
style="font-size:12.0pt;font-family:"CMU Serif"">(ms.) use
                  different algorithms (implemented in different
                  software packages), make different assumptions, and
                  their tests are conceptually quite distinct.
                  <o:p></o:p></span></li>
              <li class="MsoListParagraph"
                style="margin-left:0in;mso-list:l0 level3 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Outcomes
                  depend on many assumptions which are surrounded by
                  varying degrees of uncertainty.
                  <o:p></o:p></span></li>
            </ul>
          </ul>
        </ul>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">So
            at long last, here are my three challenge questions:<o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <ul style="margin-top:0in" type="disc">
          <li class="MsoListParagraph"
            style="margin-left:0in;mso-list:l0 level1 lfo1"><span
              style="font-size:12.0pt;font-family:"CMU Serif"">For
              the detractors of phylogenetic inference:
              <o:p></o:p></span></li>
          <ul style="margin-top:0in" type="circle">
            <li class="MsoListParagraph"
              style="margin-left:0in;mso-list:l0 level2 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">(Repeated
                from above to make sure this doesn’t get lost) Do you
                (dis)agree that if we had complete and robust knowledge
                of the phylogenies of the extant languages, we would
                perform phylogenetic tests as a matter of course when
                attempting to validate typological generalizations?<o:p></o:p></span></li>
            <li class="MsoListParagraph"
              style="margin-left:0in;mso-list:l0 level2 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Do you
                dispute that since phylogenetic testing of typological
                generalizations has become possible, typology as a field
                cannot simply ignore it and move on without it? <o:p></o:p></span></li>
          </ul>
          <li class="MsoListParagraph"
            style="margin-left:0in;mso-list:l0 level1 lfo1"><span
              style="font-size:12.0pt;font-family:"CMU Serif"">For
              the proponents of phylogenetic inference:
              <o:p></o:p></span></li>
          <ul style="margin-top:0in" type="circle">
            <li class="MsoListParagraph"
              style="margin-left:0in;mso-list:l0 level2 lfo1"><span
style="font-size:12.0pt;font-family:"CMU Serif"">Do you
                dispute that stratified sampling is statistically
                conservative and has the advantage of transparency and
                robustness vis-à-vis the underlying assumptions, so it
                is likely not a method that typology will simply move
                beyond and disregard in the future?<o:p></o:p></span></li>
          </ul>
        </ul>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif"">And
            there it. Apologies for the overlong post! – Cheers –
            Juergen<o:p></o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""><o:p> </o:p></span></p>
        <div>
          <div>
            <p class="MsoNormal"><span
style="font-size:9.0pt;font-family:Helvetica;color:black">Juergen
                Bohnemeyer (He/Him)<br>
                Professor, Department of Linguistics<br>
                University at Buffalo <br>
                <br>
                Office: 642 Baldy Hall, UB North Campus<br>
                Mailing address: 609 Baldy Hall, Buffalo, NY 14260 <br>
                Phone: (716) 645 0127 <br>
                Fax: (716) 645 3825<br>
                Email: </span><span style="font-size:11.0pt"><a
                  href="mailto:jb77@buffalo.edu"
                  title="mailto:jb77@buffalo.edu" moz-do-not-send="true"><span
style="font-size:9.0pt;font-family:Helvetica;color:#0078D4">jb77@buffalo.edu</span></a></span><span
style="font-size:9.0pt;font-family:Helvetica;color:black"><br>
                Web: </span><span style="font-size:11.0pt"><a
                  href="http://www.acsu.buffalo.edu/~jb77/"
                  title="http://www.acsu.buffalo.edu/~jb77/"
                  moz-do-not-send="true"><span
style="font-size:9.0pt;font-family:Helvetica;color:#0563C1">http://www.acsu.buffalo.edu/~jb77/</span></a></span><span
style="font-size:9.0pt;font-family:Helvetica;color:black"> <br>
                <br>
              </span><span style="font-size:11.0pt;color:black">Office
                hours Tu/Th 3:30-4:30pm in 642 Baldy or via Zoom
                (Meeting ID 585 520 2411; Passcode Hoorheh) </span><span
style="font-size:9.0pt;font-family:Helvetica;color:black"><br>
                <br>
                There’s A Crack In Everything - That’s How The Light
                Gets In <br>
                (Leonard Cohen)  </span><span style="font-size:11.0pt"><o:p></o:p></span></p>
            <p class="MsoNormal"><span style="font-size:11.0pt">-- <o:p></o:p></span></p>
          </div>
        </div>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""
            lang="DE"><o:p> </o:p></span></p>
        <p class="MsoNormal"><span
            style="font-size:12.0pt;font-family:"CMU Serif""
            lang="DE"><o:p> </o:p></span></p>
        <div id="mail-editor-reference-message-container">
          <div>
            <div
style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0in 0in 0in">
              <p class="MsoNormal"
style="mso-margin-top-alt:0in;margin-right:0in;margin-bottom:12.0pt;margin-left:.5in">
                <b><span style="font-size:12.0pt;color:black">From: </span></b><span
                  style="font-size:12.0pt;color:black">Lingtyp
                  <a class="moz-txt-link-rfc2396E" href="mailto:lingtyp-bounces@listserv.linguistlist.org"><lingtyp-bounces@listserv.linguistlist.org></a> on
                  behalf of Simon Greenhill <a class="moz-txt-link-rfc2396E" href="mailto:simon@simon.net.nz"><simon@simon.net.nz></a><br>
                  <b>Date: </b>Saturday, November 4, 2023 at 13:39<br>
                  <b>To: </b><a class="moz-txt-link-abbreviated" href="mailto:lingtyp@listserv.linguistlist.org">lingtyp@listserv.linguistlist.org</a>
                  <a class="moz-txt-link-rfc2396E" href="mailto:lingtyp@listserv.linguistlist.org"><lingtyp@listserv.linguistlist.org></a><br>
                  <b>Cc: </b>Dunn Michael
                  <a class="moz-txt-link-rfc2396E" href="mailto:michael.dunn@lingfil.uu.se"><michael.dunn@lingfil.uu.se></a>, Russell Gray
                  <a class="moz-txt-link-rfc2396E" href="mailto:russell_gray@eva.mpg.de"><russell_gray@eva.mpg.de></a><br>
                  <b>Subject: </b>[Lingtyp] Greenbergian word order
                  universals: confirmed after all<o:p></o:p></span></p>
            </div>
            <div>
              <p class="MsoNormal" style="margin-left:.5in"><span
                  style="font-size:11.0pt">Colleagues, Martin, everyone
                  else<br>
                  <br>
                  Thank you for sharing your perspectives on our 2011
                  paper. It's nice to see this still be discussed more
                  than a decade later. However, I would like to express
                  my concerns and disagreements with some of the points
                  you've raised.<br>
                  <br>
                  I'm very proud of the Dunn et al. paper for a number
                  of reasons. I'll name three.<br>
                  <br>
                  First, the paper showed that the overly simplistic
                  correlation methods that had been used to make
                  sweeping global claims were problematic. We need
                  better tools to tackle these questions, and the tools
                  we applied were one part of a better toolkit.
                  <br>
                  <br>
                  Second, it highlighted the need to understand language
                  systems in a diachronic manner. We cannot decouple
                  language typology from language history, instead we
                  need to understand how these are entangled.<br>
                  <br>
                  Third, it emphasised the way that particular
                  configurations of languages can be arrived at via
                  different routes in different families at different
                  times. This enables a much richer understanding of how
                  these particular generalisations have arisen.<br>
                  <br>
                  Have Jäger and Wähle disproved any of that? no. Maybe
                  these were not completely novel insights (Maslova’s
                  work has been mentioned which touches on a few of
                  these issues too, for example), but these ideas did
                  appear to crystallise in this paper.<br>
                  <br>
                  While it's certainly important to revisit and
                  reevaluate research findings to ensure accuracy, it is
                  crucial to approach these discussions with an
                  understanding of the scientific process. Scientific
                  paradigms evolve over time, and different studies may
                  yield varying results due to changes in methodologies,
                  data sources, and sample sizes. This doesn't
                  necessarily imply that the initial research was flawed
                  or that the authors were neglectful. In particular,
                  the tools, the data, and our understanding of how
                  languages change are substantially further advanced
                  than they were a decade ago (or, I know that *my*
                  understanding of these things is more advanced now, at
                  least). And these other papers that you mention -- and
                  many other studies -- have built upon the work we did
                  in 2011. <br>
                  <br>
                  Furthermore, I would like to caution against drawing
                  overly broad conclusions about the quality of research
                  published in high-prestige journals. The peer-review
                  process in such journals is rigorous, and while they
                  may occasionally feature sensationalist claims, this
                  doesn't diminish the overall value they contribute to
                  the scientific community. For the record, of the
                  handful of papers I've had in these journals *all*
                  have been reviewed by people I would infer to be
                  linguists based on the comments and issues they
                  raised. We did not send these papers to these journals
                  to avoid linguistic reviewers but, frankly, I've had
                  better reviews at these journals than at prominent
                  linguistics journals (and by "better" I mean more
                  rigorous, more thorough, and more critical).<br>
                  <br>
                  Finally, linguistic typology is an ongoing and
                  evolving field trying to tackle very difficult
                  problems. We need all the tools and approaches we can
                  get to solve these problems across all the levels that
                  languages operate on (from detailed language internal
                  analyses to high-level global analyses). Rather than
                  looking back and gate-keeping what is 'real’ typology
                  published in ‘real’ linguistics journals, we should
                  shift our focus forward. Typology can be a welcoming
                  and diverse community that embraces a wide range of
                  approaches, analyses, and styles. Let's look outward
                  to foster connections with other fields and
                  disciplines.
                  <br>
                  <br>
                  After all, why shouldn't linguistic typology work be
                  everywhere in science? it's certainly interesting
                  enough.
                  <br>
                  <br>
                  Simon<br>
                  <br>
                  Dr. Simon J. Greenhill<br>
                  <br>
                  Associate Professor <br>
                  <br>
                  Te Kura Mātauranga Koiora | School of Biological
                  Sciences<br>
                  Te Whare Wānanga o Tāmaki Makaurau | University of
                  Auckland<br>
                  <br>
                  Abteilung für Sprach- und Kulturevolution | Department
                  of Linguistic and Cultural Evolution<br>
                  Max-Planck-Institut für Evolutionäre Anthropologie |
                  Max Planck Institute for Evolutionary Anthropology<br>
                  <br>
                  _______________________________________________<br>
                  Lingtyp mailing list<br>
                  <a class="moz-txt-link-abbreviated" href="mailto:Lingtyp@listserv.linguistlist.org">Lingtyp@listserv.linguistlist.org</a><br>
                  <a
href="https://listserv.linguistlist.org/cgi-bin/mailman/listinfo/lingtyp"
                    moz-do-not-send="true">https://nam12.safelinks.protection.outlook.com/?url=https%3A%2F%2Flistserv.linguistlist.org%2Fcgi-bin%2Fmailman%2Flistinfo%2Flingtyp&data=05%7C01%7Cjb77%40buffalo.edu%7Cbe1181a9879e434ccc0308dbdd3319c4%7C96464a8af8ed40b199e25f6b50a20250%7C0%7C0%7C638346983806627973%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=OgjDO%2FGow%2FfFmxOIvmrWNNq1nwcsZaLzI5OMnoGjBXg%3D&reserved=0</a><o:p></o:p></span></p>
            </div>
          </div>
        </div>
      </div>
      <br>
      <fieldset class="moz-mime-attachment-header"></fieldset>
      <pre class="moz-quote-pre" wrap="">_______________________________________________
Lingtyp mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Lingtyp@listserv.linguistlist.org">Lingtyp@listserv.linguistlist.org</a>
<a class="moz-txt-link-freetext" href="https://listserv.linguistlist.org/cgi-bin/mailman/listinfo/lingtyp">https://listserv.linguistlist.org/cgi-bin/mailman/listinfo/lingtyp</a>
</pre>
    </blockquote>
    <pre class="moz-signature" cols="72">-- 
Martin Haspelmath
Max Planck Institute for Evolutionary Anthropology
Deutscher Platz 6
D-04103 Leipzig
<a class="moz-txt-link-freetext" href="https://www.eva.mpg.de/linguistic-and-cultural-evolution/staff/martin-haspelmath/">https://www.eva.mpg.de/linguistic-and-cultural-evolution/staff/martin-haspelmath/</a></pre>
  </body>
</html>