Linguistic genealogy


 Any living language is constantly changing.  If a group of people speaking the same language, call it P, breaks up into smaller groups, say A & B, without extensive communication between the groups, the changes which occur will, to some extent, be different in each group. After a period of time, perhaps as little as a generation, members of one group may be able to recognize a visiting member of the other group by his "accent", i.e., they recognize that he doesn't speak exactly the same way they do.  At this point we have the beginnings of a dialect situation.  And when these dialects A & B at a later point in time are so different that speakers of A are unable to communicate with speakers of B, we say that A and B are related but different languages, descended from a common "parent", language P.  Where historical records are available, genetic relationships such as that between hypothetical languages A & B are easily documented.

But historical documentation in the Americas goes back only a short way.  How can we establish that two or more native languages are genetically related, i.e. descended from a common,  unattested, proto-language?

Before discussing this, it is important to clarify certain things about our model of linguistic relationships.  In saying two languages are 'descended from a common parent', we are using a kinship metaphor.  Thus the model we end up with is a 'family tree' model. Models can be misleading unless we are fully aware of their limitations. In this case, we must remember that the 'parent' language doesn't "die"; it has changed into its 'daughter' languages.  Furthermore, the model reinforces the convenient fictions that groups such as A and B were homogeneous speech communities, separated suddenly, and had little if any linguistic effect on each other after separation, whereas in actuality none of these is likely to be true.

Establishing genetic relationships

 Returning to our questions of how we can establish pre-history linguistic relationships, we consider words from some current native languages.

Looking over vocabulary of Zuni and Southern Tiwa languages of New Mexico, one finds the following words:
    Zuni         wa:kashi      'cattle/bull/cow'
    So. Tiwa     bakade        'cow'
These words are similar enough that we might suspect that they are descended from a common parent; i.e., we might hypothesize that they are cognates which developed from a single word of a proto-language which is the progenitor of both Zuni and So. Tiwa.  Finding the next pair of words could reinforce that idea:
    Zuni           kolta      'law court'
    So. Tiwa     kurti      'law court'
 Such a relationship between these languages would seem reasonable, since these two groups of people have similar cultures and are relatively close geographically (about 200 kilometers).  But we require much more evidence than this.  The similarity seen in the pairs of words for 'cow' and for 'court' could have explanations other than genetic relationship between the languages.  In the case of the first pair, knowledge that cattle were not indigenous to North America should make us suspicious, and comparison with Spanish vaca 'cow' and vacas 'cows' indicates that the similarity between the Zuni and So. Tiwa words for cow is due to borrowing by both languages from Spanish.  Similarily the words for 'court' in both languages are suspiciously like the English word, so borrowing again seems likely.  So we need a way to rule out borrowing between languages, either from one another or from a third language, before we can accept similarity as evidence for genetic relationship.

There is another possible source of similarities such as those shown above:  pure coincidence.  This is evidently the case for the following words from Eskimo and Blackfoot:
    Eskimo       akisi       'bed'
    Blackfoot   akihsii    'bed'
 We want a method to rule out such chance similarities as well, in our attempt to reconstruct linguistic history.

The method to be illustrated here was developed by philologists and linguists working with Indo European languages, where there were many historical records to serve as checkpoints to confirm the adequacy of the method.  It is based on the premise that sound change is regular; i.e., that if a given proto-language sound, e.g. t, in a given word turns up in a daughter language as another sound, e.g. s, then every word of the proto-language containing t will have a corresponding s in the daughter language. Any exceptions must be accounted for by additional explanations such as the presence of neighboring sounds in the exceptional word that will affect the change.

To illustrate why regularity of sound change enables us to establish that languages are genetically related, we now compare some words of our hypothetical languages A and B to their earlier form in the parent language P.  See table 1.  First, compare A with P.

 
                        TABLE 1.
 
      meaning         P         A         B   
 
     'man'           ati       ase         oci
     'leg'             tin       sen        cin
     'owl'           otip      osep      ucip
     'river'         ipen      epen      ipen
     'moon'         etit      esis       ecic

 In the word for 'man', the t of P shows up as s in A.  If sound change is regular, we should find s in every word of A if that word had t in P.  And we see this to be true in the words meaning 'leg', 'owl', and 'moon'.  We can state this as a "sound law": P t becomes A s.  Comparing now B with P, we find that t regularly shows up as c in B, giving us sound law P t becomes B c.   We can also state other sound laws for the development of A and B from P as follows:  P a becomes A a; P a becomes B o; P i becomes A e; P o becomes B u; etc.  Table 2 summarizes these.

TABLE 2.
 
    P    A    B
    t      s     c
    a     a     o
    i      e     i
    n     n    n
    o     o    u
    p     p    p
    e     e     e
 
 Now, suppose we had no record of language P, but only data from A and B.  Comparing only the A and B words of Table 1 we could discover that wherever A has s, B has c; i.e. there is a regular correspondence of A s to B c.  Similarily, wherever A has a, B has o.  If we summarized all the correspondences, and put them in a table, it would look like the A and B columns of Table 2.  We could be sure that such regularity of correspondence was not due to chance, and if it holds up when a larger sample of vocabulary is used, then it is also unlikely that the similarity between these languages is due to borrowing.

We are ready to apply the method to some real data.  To make things more interesting, we will use five languages rather than just two. Because of space limitations, we will use only eight words from each language, whereas we would normally want a much larger sample.  The sets of words with nearly identical meanings and similarity of shape are listed in Table 3.  Looking first at the words for  'my sibling in law', 'my louse', and 'three' we observe that they all start with n in all five languages.  This qualifies as a regular correspondence; however, by itself it proves nothing. But if these same words contain other regular correspondences, and all of these correspondences are found in other words, then we have established a genetic relationship.  When we look for other regular correspondences in our first three words, we find that the third syllable of the first word has m in all five languages; and 'my  louse' also has an m in all five languages.  (This matching of m's can also be seen in a number of other words).  Looking at the first vowel of 'my louse' and 'three' we see the following correspondence set in both columns:  e;i;e;i;a.  This same set is found in the second syllable of 'my sibling in law', in the first syllable of 'ten', and in the third syllable of the stem in 'he trusts'.

 
                   Table 3
 
                       man            dog              ten           he gives to him
 
Fox               ineniwa     anemwa      meta:swi        mi:ne:wa
Cree             iyiniw         atim            mita:taht        miye:w
Menomini   enE:niw      anE:m         meta:tah         me:nE:w
Ojibwa         inini           anim           minta:sswa     mi:na:t
Cheyenne   hetáne       hotáme          mahtóhto     é-méto
 
               my sibl. in law        my louse             three            he trusts
 
Fox             ni:nemwa            netehkoma         neswi          ahpe:nemowa
Cree            ni:tim                  nitihkom            nisto              aspe:yimow
Menomini   ne:nem                netE:hkom         ne'niw          ahpE:nemow
Ojibwa        ni:nim                 nintikkom          nisswi          appe:nimo
Cheyenne   nétame                nahtáeme            na'heh        e-ne'étame
 

The regularity of the correspondences we have pointed out thus far is enough to make a good case for a genetic relationship here.  But since each correspondence we have discussed is a hypothesis that a certain (unknown) sound of the unrecorded parent language ended up as a certain sound in each of these languages, we must deal with certain apparent counterexamples to the regularity of our correspondences. Notice that in 'man' we find an <n;y;n;n;t> set, in addition to the <n;n;n;n;n> we observed earlier.  Does this mean that either our hypothesis is wrong, or that we must conclude that the set of words for 'man' are not cognate; i.e. not descended from a common proto-word?  It evidently means neither of these things, because not only does 'man' show other correspondences that are regular, but the <n;y;n;n;t> correspondence set is found also in 'he gives to him' and 'he trusts'. So this is simply a different set of regular correspondences, presumably because this set of sounds reflects a different proto-language sound.  The same is true of the <n;t;n;n;t> set seen in 'dog' and 'my sibling in law'.  We can symbolize these unattested sounds of the proto-language as L and D, being aware that these are simply symbols for the correspondences, and do not necessarily reflect the actual sounds of the proto-language.

There is much more we could say about correspondences in this set of words, but this much should suffice to illustrate how genetic relationships can be established.


[Back to Algonquian features of Blackfoot]
[Back to Site Map]