The Interpretation of Relationships among Quechua Dialects Author(s): Joseph E. Grimes Reviewed work(s): Source: Oceanic Linguistics Special Publications, No. 20, For Gordon H. Fairbanks (1985), pp. 271-284 Published by: University of Hawai'i Press Stable URL: http://www.jstor.org/stable/20006728 . Accessed: 18/03/2013 09:01 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. . University of Hawai'i Press is collaborating with JSTOR to digitize, preserve and extend access to Oceanic Linguistics Special Publications. http://www.jstor.org This content downloaded on Mon, 18 Mar 2013 09:01:17 AM All use subject to JSTOR Terms and Conditions THE INTERPRETATION OF RELATIONSHIPS AMONG QUECHUA DIALECTS by Joseph E. Grimes Cornell University Summer Institute of Linguistics ABSTRACT In historical linguistics it is desirable to keep a balance between the details worked out by careful attention to methodology on the one hand, and the overall picture that emerges from those details on the other. Though the historical study of the Quechua dialects is not far advanced, there are measures of the mass effect of all known factors that give us the shape of the picture that is emerging there. This paper begins from lists of cognates compiled by Alfredo Torero and applies two grouping algorithms to his judgments. The first is in terms of simple links among the dialects, and the second in terms of optimal networks of dialects at different levels of association. Both results suggest that the different branches of Quechua A may be as distant from each other as any of them is from Quechua B. QUECHUA DIALECTS There are several ways to find out about patterns of intercommunication among speech communities and the situations that led up to them. Part of the evidence is purely linguistic; another part is sociolinguistic. The linguistic evidence may incorporate classic methods taken from dialectology in which an account is given of the geographic or social spread of some selection of phonetic, phonemic, lexical, or grammatical features. This type of study pays special attention to areas where a number of features agree independently in dividing one area from another. Another kind of evidence is at the level of abstraction of historical linguistics, that is, it groups the speech of various communities according to the way some of them but not others have participated in the processes of change. When it comes to forming a picture of the total pattern of linguistic divergence, these approaches show a common weakness. Since they build on observations of selected traits, it is difficult to assign a weight to the effect any particular trait has in regard to maintenance or failure of communication. One reason for this is that the contribution of each trait is relative not only to the other traits studied, but also to those traits that do not happen to be selected for study. It is therefore useful to look at the specific features of change against a background of measures of change that show the joint effect of all factors working together. This content downloaded on Mon, 18 Mar 2013 09:01:17 AM All use subject to JSTOR Terms and Conditions 272 MASS MEASURES Another kind of research has concentrated almost exclusively on the linguistic and sociolinguistic evidence that gives an overview of language groupings at the cost of not being able to identify any of the particular traits that are responsible for the total effect. These mass measures are nearly always quantitative. They include lexicostatistics (Swadesh 1950, 1955), intelligibility tests (Voegelin and Harris 1951, Grimes 1974, Casad 1 974, Simons 1 979), dialect adaptation tests (Weber and Mann 1982), and the quantification of phonological divergence which Simons has labelled phonostatistics (Grimes and Agard 1959, Grimes 1964). Being expressed in quantitative form, of course, is no guarantee that measures of the second type are accurate. Lexicostatistics especially has been criticized for the looseness of all its assumptions. Nevertheless, even though they are inadequate in many respects. It is the mass measures that offer the most for formulating an overall perspective on the interrelationships among the members of a linguistic family. What they provide is a general layout within which it is possible to examine the contribution of specific details without losing the overall sense of proportion. What is sometimes overlooked is that a set of numbers that expresses the divergence among languages Is derived from a set of phenomena that are by nature interrelated. As a result, no number in the set means anything by itself. Every one is a component of a matrix, and the matrix as a mathematical entity is what leads to the interpretation of its parts. It expresses the total system of interrelationships. That is why the interpretation of quantitative measures of divergence is not just a matter of finding numbers that are noticeably larger or smaller than others. Each number has to be interpreted through its relationship to the whole matrix. SINGLE LINKG ROUPINGS This paper presents two matrix based procedures that help a linguist get a global view of language relationships based on mass measures. The first looks at pairs of linguistic entities - I refer to them as 'dialects' without entering into the question of which are separate languages and which are not. It takes the two most closely related ones in the matrix and assigns them to a new fictitious entity that relates to the rest of the matrix by the closer of each of the two figures for the entities of which it is composed. In other words, if M(i,j) is the smallest number in a matrix M in which close relationships are represented by small numbers, then the two rows M(i,*) and M(j,*) are replaced by a single row M(i',*) in which each cell M0',k) is the lesser of M(i?k) or M(j,k) for each column k. The same process is repeated for the two columns M(*,i) and M(*,j). The original value M(i,j) is taken as the branching level of a tree like the one in Figure 1, whose descendents are dialects / and j if they are terminal, or else fictitious entities like /" if they are nonterminal. The fictitious entity replaces the two rows and columns from which it was formed, and the process continues until the whole matrix is reduced to a single fictitious entity. This first procedure, called to my attention by Louis L. McQuitty (1964). has This content downloaded on Mon, 18 Mar 2013 09:01:17 AM All use subject to JSTOR Terms and Conditions 273 been called a process of grouping by single links, because each element is added into a group on the basis of a single factor, namely its closest relationship among the other elements. Figure 1 shows the results of applying this process to information about Quechua dialects from South America that will be described in due course. The nodes of Figure 1 are identified in Table 1, and the matrix is in Table 2. This content downloaded on Mon, 18 Mar 2013 09:01:17 AM All use subject to JSTOR Terms and Conditions 274 Figure 1. Quechua dialects after Torero 1970, single link grouping, with Torero's classification on the left. This content downloaded on Mon, 18 Mar 2013 09:01:17 AM All use subject to JSTOR Terms and Conditions 275 Figure 2- Quechua dialects \ \ ViX ^M^T,sl^ J/Wi^^mS, after Torero 1970, \\ /// G^yv^V /l\?\ grouped into \\\\\\/<-[^/(^ ??s/s'Z ^yyyv/?^^^C\/?)\?\ 3// / 11 networks at \s\SSSS^?^^y^/r^\ S / th.. reshol.d s -,j 11, ,, \\N\\xXi?^X/ Xj? /^-^^ (vMun// /I This content downloaded on Mon, 18 Mar 2013 09:01:17 AM All use subject to JSTOR Terms and Conditions 276 Table 1. Symbols used to represent Quechua dialects in Torero 1970, with symbols for his classification. Ali Alis, Lima, Peru: I Anj Andajes, Lima, Pero: I Aye Ayacucho, Ayacucho, Peru: IIC Car Caras, Ancash, Pero: I Ccr Cacra, Lima, Peru: I Chg Chongos Bajo, Junin, Peru: I Chp Chachapoyas, Amazonas, Peru: IIB Chq Chiqui?n, Ancash, Peru: I Chu Chuquisaca, Bolivia: IIC Cjm Cajamarca, Cajamarca, Peru: IIA Cjt Cajatambo, Lima, Peru: I Cor Corongo, Ancash, Per?: I Cuz Cuzco, Cuzco, Pero: IIC Fer Ferre?afe, Lambayeque, Peru: IIA Gra Grau, Apurlmac, Pera: IIC Hgs Huang?scar, Lima, Per?: I Hri Huari, Ancash, Per?: I Ini La Uni?n, Hu?nuco, Per?: I Jau Jauja, Junin, Per?: I Lam Lamas, San Martin, Per?: IIB Lar Laraos, Lima, Peru: IIA Lin Lincha, Lima, Per?: IIA Mnz Monz?n, Hu?nuco, Per?: I Mun Mu?ecas, La Paz, Bolivia: IIC Ocr Ocros, Ancash, Per?: I Pan Panao, Hu?nuco, Per?: I Pch Pichincha, Ecuador: IIB Per Pacaraos, Lima, Per?: IIA Pot Potos?, Bolivia: IIC Puq Puquio, Ayacucho, Per?: IIC SdE Santiago del Estero, Argentina: IIC Sih Sihuas, Ancash, Per?: I Sur Surcubamba, Huancavelica, Per?: IIC Tap T?puc, Pasco, Per?: I Tnt Tantar?, Huancavelica, Per?: I Trm Tarma, Jun?n, Per?: I Ule Ulcumayo, Jun?n, Per?: I This content downloaded on Mon, 18 Mar 2013 09:01:17 AM All use subject to JSTOR Terms and Conditions 277 *** Cor Sih Hri Mnz Pan Car Ocr Chq Uni Cjt Anj Tap Ule Trm Jau Ali Cor 0 7 24 35 27 '31 42 39 21 31 35 31 24 27 50 54 Sih 7 0 17 27 27 27 39 39 21 31 35 31 17 27 54 50 Hri 24 17 0 24 31 31 31 42 31 35 46 39 31 31 42 42 Mnz 35 27 24 0 21 35 39 35 21 24 42 46 27 24 50 46 Pan 27 27 31 21 0 35 39 35 10 21 31 35 31 24 46 54 Car 31 27 31 35 35 0 24 31 24 21 31 39 31 17 39 42 Ocr 42 39 31 39 39 24 0 24 27 39 50 46 39 39 31 39 Chq 39 39 42 35 35 31 24 O 27 27 35 42 39 35 42 50 Uni 21 21 31 21 10 24 27 27 O 17 24 24 21 21 31 35 Cjt 31 31 35 24 21 21 39 27 17 O 14 39 35 14 35 50 Anj 35 35 46 42 31 31 50 35 24 14 O 42 39 24 46 61 Tap 31 31 39 46 35 39 46 42 24 39 42 O 27 35 46 54 Ule 24 17 31 27 31 31 39 39 21 35 39 27 O 24 39 35 Trm 27 27 31 24 24 17 39 35 21 14 24 35 24 O 24 35 Jau 50 54 42 50 46 39 31 42 31 35 46 46 39 24 O 31 Ali 54 50 42 46 54 42 39 50 35 50 61 54 35 35 31 O Chg 58 61 54 50 50 54 50 50 39 50 61 53 42 35 35 27 Ccr 42 42 39 39 46 46 42 54 35 42 54 54 39 31 31 24 Hgs 61 61 58 54 50 54 54 66 42 50 58 70 54 35 39 27 Tnt 82 82 82 78 70 82 82 82 61 74 78 78 74 58 58 46 Per 42 39 31 46 42 31 42 50 39 31 39 54 54 31 46 46 Per 46 46 58 66 61 70 70 70 46 66 70 54 50 54 58 54 Cjm 50 50 58 70 70 58 70 70 54 58 61 70 54 42 46 50 Lar 58 50 42 39 46 46 46 58 39 46 54 54 39 31 31 17 Lin 66 66 54 54 54 58 54 66 46 54 61 58 53 39 35 27 Chp 86 78 82 82 86 86 86 95 70 86 95 86 78 78 82 58 Pch 95 86 86 86 95 82 91 95 73 82 95 104 82 74 74 58 Lam 104 95 86 86 95 100 100 109 86 104 104 95 91 86 91 78 Sur 86 74 74 70 78 82 86 86 66 78 78 86 66 61 70 39 Ayc 91 82 74 78 86 86 82 95 70 86 86 82 74 74 66 46 Puq 86 78 74 74 82 86 86 95 70 78 91 74 70 66 61 42 Gra 91 78 82 82 86 91 70 74 74 82 100 91 86 78 74 61 Cuz 82 78 70 74 82 82 74 86 66 78 91 78 78 78 70 50 Mun 91 82 78 78 82 86 74 82 70 82 95 95 86 82 74 54 Pot 95 86 78 82 91 95 70 86 78 91 104 95 86 86 82 66 Chu 95 86 82 78 82 95 82 86 78 91 104 100 86 82 86 61 SdE 100 91 78 86 82 95 70 36 82 95 104 100 95 91 74 70 Table 2a. Matrix of relationships among Quechua dialects. Cognate percentages transformed into minimum centuries X 10, after Torero 1970. This content downloaded on Mon, 18 Mar 2013 09:01:17 AM All use subject to JSTOR Terms and Conditions 278 *** Chg Ccr Hgs Tnt Per Per Cjm Lar Lin Chp Pch Lam Sur Aye Puq Gra Cor 53 42 61 82 42 46 50 58 66 36 95 104 86 91 86 91 Sih 61 42 61 82 39 46 50 50 66 78 86 95 74 32 78 78 Hri 54 39 58 82 31 58 58 42 54 82 86 86 74 74 74 82 Mnz 50 39 54 78 46 66 70 39 54 82 86 86 70 73 74 82 Pan 50 46 50 70 42 61 70 46 54 86 95 95 78 86 82 86 Car 54 46 54 82 31 70 58 46 58 86 82 100 82 86 86 91 Ocr 50 42 54 82 42 70 70 46 54 86 91 100 86 82 86 70 Chq 50 54 66 82 50 70 70 53 66 95 95 109 86 95 95 74 Uni 39 35 42 61 39 46 54 39 46 70 78 86 66 70 70 74 Cjt 50 42 50 74 31 66 58 46 54 86 82 104 73 86 78 82 Anj 61 54 58 78 39 70 61 54 61 95 95 104 78 86 91 100 Tap 58 54 70 78 54 54 70 54 53 86 104 95 86 82 74 91 Ulc 42 39 54 74 54 50 54 39 53 78 82 91 66 74 70 86 Trm 35 31 35 58 31 54 42 31 39 78 74 86 61 74 66 73 Jau 35 31 39 58 46 58 46 31 35 82 74 91 70 66 61 74 Ali 27 24 27 46 46 54 50 17 27 58 53 78 39 46 42 61 Chg 0 35 35 50 58 61 54 27 31 78 61 82 50 53 58 70 Ccr 35 0 24 42 58 54 58 21 24 78 74 86 58 50 58 66 Hgs 35 24 0 21 61 70 66 21 10 61 58 78 31 35 35 39 Tnt 50 42 21 0 74 78 66 35 24 46 54 54 17 24 17 27 Per 58 58 61 74 0 54 42 46 58 78 70 95 70 82 74 78 Per 61 54 70 78 54 0 27 58 66 66 82 86 78 78 74 100 Cjm 54 58 66 66 42 27 0 54 61 61 66 82 70 70 74 86 Lar 27 21 21 35 46 58 54 0 10 58 54 70 35 35 35 54 Lin 31 24 10 24 58 66 61 10 0 66 61 78 35 31 31 46 Chp 78 78 61 46 78 66 61 58 66 0 35 10 39 54 31 58 Pch 61 74 58 54 70 82 66 54 61 35 0 42 54 61 61 61 Lam 82 86 78 54 95 86 82 70 73 10 42 0 46 61 42 74 Sur 50 58 31 17 70 78 70 35 35 39 54 46 0 17 10 35 Aye 58 50 35 24 82 78 70 35 31 54 61 61 17 0 21 31 Puq 53 58 35 17 74 74 74 35 31 31 61 42 10 21 0 39 Gra 70 66 39 27 78 100 86 54 46 58 61 74 35 31 39 0 Cuz 61 54 39 27 74 91 78 39 35 46 46 66 31 21 27 17 Mun 66 61 42 42 86 95 91 46 46 54 54 74 42 39 46 24 Pot 74 74 58 46 86 95 91 58 54 50 39 54 42 42 42 35 Chu 66 74 50 39 82 91 86 58 54 42 39 53 35 42 42 35 SdE 74 70 54 35 95 109 100 53 53 50 46 54 46 54 46 35 Table 2b. Matrix of relationships among Quechua dialects. Cognate percentages transformed into minimum centuries X 10, after Torero 1970. This content downloaded on Mon, 18 Mar 2013 09:01:17 AM All use subject to JSTOR Terms and Conditions 279 *** Cuz Mun Pot Chu SdE Cor 82 91 95 95 100 Sih 78 82 36 86 91 Hri 70 78 78 82 78 Mnz 74 78 82 78 86 Pan 82 82 91 82 82 Car 82 86 95 95 95 Ocr 74 74 70 82 70 Chq 86 82 86 86 86 Uni 66 70 78 78 82 Cjt 78 82 91 91 95 Anj 91 95 104 104 104 Tap 78 95 95 100 100 Ule 78 86 86 86 95 Trm 78 82 86 82 91 Jau 70 74 82 86 74 Ali 50 54 66 61 70 Chg 61 66 74 66 74 Ccr 54 61 74 74 70 Hgs 39 42 58 50 54 Tnt 27 42 46 39 35 Per 74 86 86 82 95 Per 91 95 95 91 109 Cjm 78 91 91 36 100 Lar 39 46 58 58 58 Lin 35 46 54 54 58 Chp 46 54 50 42 50 Pch 46 54 39 39 46 Lam 66 74 54 58 54 Sur 31 42 42 35 46 Ayc 21 39 42 42 54 Puq 27 46 42 42 46 Gra 17 24 35 35 35 Cuz O 21 24 24 39 Mun 21 O 24 21 35 Pot 24 24 0 10 24 Chu 24 21 10 0 27 SdE 39 35 24 27 0 Table 2c. Matrix of relationships among Quechua dialects. Cognate percentages transformed into minimum centuries X 10, after Torero 1970. This content downloaded on Mon, 18 Mar 2013 09:01:17 AM All use subject to JSTOR Terms and Conditions