You are on page 1of 65

MA Thesis

Social deixis and classifiers

Supervisors:
Prof. Dr. P.C. Muysken and Dr. E.I. Crevels

Student:
O. Krasnoukhova, 0435759

June 2007
Radboud University Nijmegen
Acknowledgements

My foremost gratitude goes to my supervisors, Prof. Dr. Pieter Muysken and Dr. Mily
Crevels for giving me the opportunity to explore and learn, and for generously sharing
their knowledge and experience.
I am indebted to Prof. Dr. P. Muysken for his insight on the approach of the topic
and for providing me with valuable feedback and guidance.
I am deeply grateful to Dr. M. Crevels for her dedication and support throughout
the course of the thesis, for teaching me how to work with data and for numerous
valuable suggestions that have improved the study greatly. My sincere thanks to her
for reading the drafts so thoroughly and for correcting them with the skill which I
admire.
Special thanks go to Prof. Dr. S.C. Levinson, Prof. Dr. A.Y. Aikhenvald, Prof. Dr. N.
England, Prof. Dr. G. Senft, who have taken the time to answer my questions, and to
Dr. N.J. Enfield for the comments, advice and encouragement.
My warmest thanks go to Wil Strijbos, Ruti Vardi, Isabelle and Jeroen Bendermacher,
whose friendship, encouragement and great sense of humour often helped to keep
afloat. Thanks from all my heart to Anna and Janne Östlund for letting me be their
‘rysska dotter’.
Never enough thanks to my dear parents for their love and for making impossible
possible. I dedicate this work to them.
Contents

1. Introduction 1

2. Social deixis 3
2.1. The notion of social deixis 3
2.2. Examples of social deixis. Case studies 4
2.2.1. European languages: T/V pronoun phenomenon 4
2.2.2. Sinhalese: pronoun system 6
2.2.3. Guugu Yimidhirr: avoidance behaviour 7
2.2.4. Japanese: honorifics 9
2.2.5. Samoan: respect vocabulary 13
2.3. Summary 13

3. Classifiers 14
3.1. Systems of nominal classification and the place that classifiers hold among them 14
3.2. Types of classifiers 16
3.2.1. Numeral classifiers 16
3.2.2. Noun classifiers 18
3.2.3. Genitive classifiers 19
3.2.4. Verbal classifiers 23
3.2.5. Locative classifiers 24
3.2.6. Deictic classifiers 24
3.3. General remarks on classifiers 25

4. Classifier systems of social interaction 26


4.1. Genitive classifiers. Social status: Pohnpeian 27
4.1.1. Pohnpeian 27
4.1.2. Arguments for the hypothesis 29
4.2. Noun classifiers. Age and kinship: Jacaltec, Mam and Akateko 32
4.2.1. Jacaltec 32
4.2.2. Mam 35
4.2.3. Akateko 36
4.2.4. Arguments for the hypothesis 38
4.3. Numeral classifiers. Family membership: Yi branch languages 41
4.3.1. Akha and Hani (Southern Yi languages) 42
4.3.2. Nasu and Nosu (Northern Yi languages) 44
4.3.3. Lisu (Central Yi Language) 45
4.3.4. Arguments for the hypothesis 46
4.4. Numeral classifiers. Religious address forms: Burmese and Thai 47
4.4.1. Burmese 47
4.4.2. Thai 48
4.4.3. Arguments for the hypothesis 50
5. Conclusion and discussion 51
References 56
List of tables
nd
Table 1 The 2 person pronoun and social variables in Sinhalese 6

Table 2 Genitive classifiers in common, exaltive honorific and humiliative honorific 28

Table 3 Genitive classifiers in common and humiliative honorific 29

Table 4 Jacaltec noun classifiers. System of social interaction 33

Table 5 Noun classifiers in Mam 35

Table 6 Noun classifiers in the Akateko speaking areas 37

Table 7 Kin terms and family group classifiers in Akha 43

Table 8 Family group classifiers in Nasu 44

Table 9 Additional family group classifiers in Nasu 45

Table 10 Numeral classifiers in Burmese categorizing human beings 48

List of figures

Figure 1 Use of the T/V pronouns along two dimensions: power and solidarity 5

Figure 2 Continuum of nominal classification systems 14

Figure 3 Jacaltec classifiers of social interaction 35

Figure 4 Mam noun classifiers along the dimensions of age, respect and sex 36

Figure 5 Internal organization of the classifiers of social 38

Figure 6 Numeral classifiers in Thai used for animate beings 49


List of abbreviations

ABS absolutive
AG agent
AN animate
AP antipassive suffix
CL noun class
CLF classifier
COM comitative
COP copula
D dual
DEM demonstrative
DAT dative
DISRESP disrespectful
ENCL enclitic
EXAL exaltive
EXCL exclusive
GNR generic
F female
FGCL family group classifier
GEN genitive
HUM human
HON honorific
HML humiliative
IMP imperative
INAN inanimate
INCL inclusive
LOC locative
M masculine
NUM numeral
NEUT neutral
NKIN non-kin
NPST non-past
PL plural
POL polite
POSS possessive
POSSR possessor
PROGR progressive
PROX proximal
PRTCL particle
REL relative
RESP respectful
RN relational noun
RP repeater
SG singular
SUB subordinate clause
TOP topic
1D one-dimensional
1 first person
2 second person
3 third person
1. Introduction
The definitional property of classifiers to classify referents into particular categories
has provoked a whole number of studies addressing the question of “classifiers as a
reflection of mind” (e.g. Lakoff 1986, Craig 1986a,b, Aikhenvald 2000, Senft 2000).
The transparency of such categories has been shown to vary from language to
language: categories may include remarkably similar prototypical members,
exhibiting strong semantic motivation, or they may resemble a container combining
very heterogeneous items. One of the numerous examples of the latter, which has
become classic already, is the categorization in Dyirbal described by Dixon (1982)
and discussed by Lakoff (1986), where one of the four categories includes ‘women,
water, fire, fighting, dogs, scorpions and the hairy mary grub’ to name a few. All
systems of classification, though, share a common set of semantic principles of
categorization, main of which are animacy and physical properties. Among the
semantic features of classification, one can establish an implicational scale “which
states that linguistic classifications mark humanness and animacy first, then shape,
then use and consistency” (Craig 1986a:5). Some languages show categorization of
nouns according to only one of these parameters, some exhibit the full range. Besides,
a whole number of languages is attested to have classifiers where categorization of
animate referents is not limited to only humanness and animacy. Such languages are
reported to have sets of classifiers which go further and categorize their animate
referents according to various parameters typical for only animates. These can include
age, sex, kin relationship, religious affinity, spiritual and supernatural qualities. The
parameter that seems to be encoded by classifiers most often is status, “the
hierarchical ranking of individuals along a dimension of social prestige” (Foley
1997:309). To such sets of classifiers I prefer to refer with the cover term of “social
interaction” used by Craig (1986b, following Denny 1976) in her description of
Jacaltec classifiers. These sets of classifiers of social interaction constitute the major
interest of the present thesis.
One of the characteristic features of classifiers is their ability to add semantic
content to the noun phrase in which they occur (Dixon 1982, 1986, cited in Grinevald
2000:61) (see Section 3). Besides, in contrast to gender which is obligatory and
unchangeable, the use and the choice of a classifier can be partly determined by the
speaker. Thus, a referent noun can be at times assigned by a speaker to more than one
category, depending on what property of the referent noun he wishes to emphasize.
Classifiers of social interaction have been attested to have these qualities as well.
Reclassification possibilities and the dynamism of some classifier systems of social
interaction have demonstrated that a speaker can express various attitudes towards a
referent ‘playing’ with classifiers. The stylistic reclassifications within the system of
human classification can take place along such dimensions as status, age, kinship,
or/and religious affinity. Classifier morphemes are therefore one of the linguistic
means to communicate social relation and various attitudes: from worship, respect or
solidarity to social distance or personal dislike. However, metaphoric usage and
reclassification can take place not only within the paradigm of classifiers used with
human referents, but can be extended to the entire classifier paradigm present in the

1
given language. Such extensions allow, for example, a human being to be classified as
a non-human, or an object. For instance, in Khmer (Mon-Khmer), criminal types are
optionally classified as people (Maspero 1915 and Headley et al. 1977, cited in
Adams 1986:244) and, in Bahnar (Mon-Khmer), children have been reported to be
classified as ‘seed, grain’, probably in order to emphasize their small size and
growing nature (Adams 1986:243).
The standard categorization process, however, reflects and is determined by
certain social realities in which a speech act occurs and by the social identities of the
referent and the speaker. Categorization according to status or kinship is always
relative, since it is made from the standpoint of one of the speech act participants. It
can be argued therefore that classifiers used with human referents show properties of
socially deictic items. This statement, defined in the following, I take as the
hypothesis to be investigated in the present thesis. Thus, the hypothesis is: systems of
classifiers used for categorization of human referents can have the function of social
deixis.
In what follows, I consider such phenomena as social deixis and classifiers,
and attempt to find arguments in support or against the hypothesis. For the analysis
several groups of classifier languages have been selected, namely:
(i) Pohnpeian (Micronesian);
(ii) Jacaltec, Mam and Akateko (Mayan);
(iii) Yi branch languages (Tibeto-Burman);
(iv) Burmese (Sino-Tibetan) and Thai (Tai).
The selection of these language families is motivated by the following:
(a) each of the classifier inventories of the selected languages has a dominant
social parameter according to which human referents are classified, namely:
status (genitive classifiers in Pohnpeian); kinship, age and sex (noun
classifiers in Jacaltec and Mam); family membership (numeral classifiers in
Yi branch languages), and religious affinity (numeral classifiers in Burmese
and Thai);
(b) the languages demonstrate that at least numeral, noun and genitive classifier
types can be engaged in the encoding of information on social identity;
(c) the languages have relatively large inventories of classifiers used with human
referents.
Besides, it should be mentioned that information on classifier systems of social
interaction is limited to a small number of languages.
To approach the question gradually, the phenomenon of social deixis and some forms
of it are considered first (Section 2). Section 3 deals with systems of nominal
classification and the place that classifiers hold among them. Section 4 looks at the
above mentioned classifier languages and their inventories of classifiers used with
human referents. Conclusion and discussion are outlined in Section 5.

2
2. Social deixis
2.1 The notion of social deixis
Deictic items are those “properties of utterances which are determined by, and which
are interpreted by knowing, certain aspects of the communication act in which the
utterances in question can play a role” (Fillmore 1975:39). Deixis includes the
following categories: person deixis covering identities of interlocutor in
communication, for instance personal pronouns; place deixis referring to the place in
which interlocutors are located, for instance there, here; time deixis encoding the time
at which a communication act takes place or the time at which the message is sent or
received, for example now, then; and discourse deixis, which Fillmore (1975:70)
characterizes as “lexical or grammatical elements which indicate or otherwise refer to
some portion or aspect of the ongoing discourse”, for instance, words and expressions
like the former, the latter, in the following paragraph.
Levinson (1979:206, referring to Lyons 1968:280) notes that these categories
of deixis can be extended by adding one more category, the one of social deixis,
which Fillmore (1975:76) defines as “the study of that aspect of sentences which
reflect or establish or are determined by certain realities of the social situation in
which the speech act occurs”. Under the cover term of social deixis Fillmore
(1975:76) includes the following linguistic phenomena:
• devices for person marking, e.g. pronouns;
• the various ways of separating speech levels;
• distinctions in utterances of various types which are dependent on certain
properties of the speech act participants;
• “the various ways in which names, titles, and kinship terms vary in form and
usage according to the relationships among the speaker, the addressee, the
audience and the person referred to”;
• linguistic performance which can count as social acts, such as insults,
greetings, expressions of gratitude;
• linguistic performances which can accompany social acts, such as ‘there you
go’, etc.;
• and, finally, the various linguistic devices that helps a speaker to “establish
and maintain a deictic anchoring with a given addressee”.
As can be seen above, the term social deixis defined by Fillmore (1975) absorbs
various phenomena. Levinson (1979:206) proposes to “restrict the term to those
aspects of language structure that are anchored to the social identities of participants
(including bystanders) in the speech event, or to relations between them, or to
relations between them and other referents”. Besides, various aspects of language
depending on these relations can be regarded as relevant to social deixis “in so far as
they are grammaticalized”.
Further, Levinson (1979:207) distinguishes two kinds of social dimension that
get encoded in languages: relational and absolute. The former includes relations
between a) speaker and referent; b) speaker and addressee; c) speaker and bystander;
and d) speaker and setting. The latter includes forms which are reserved for two

3
specific kinds of speech act participants, namely authorized speakers and authorized
recipients. Levinson (1979:207) exemplifies the forms which can be employed by a
certain kind of speakers by the Thai polite particles khráb which is used only by male
speakers and khá, used exclusively by women. The similar situation can be observed
in speech of Japanese men and women (see Section 3.2.4). Both gender groups can be
referred to as authorized speakers in such cases. Authorized recipients, on the other
hand, are forms denoting titles of address, pronouns differing not only with sex of
referent but also with sex of addressee, and other forms used for a particular kind of
recipients.
As the definition of social deixis suggests, the milieu of existence of social deixis is a
social context. Manning (2001) considers the phenomenon of social deixis from a
qualitatively different angle. He argues that in the studies so far, social context has
been interpreted implicitly as the concrete speech situation. The standpoint of
Manning is that there are forms of social deixis where social context should rather be
interpreted as ‘sociocultural context’. In other words, there are forms of deixis that
“make reference to relatively perduring “social” rather than emergent “situational”
relations”; and these are the forms for which he reserves the term ‘social deixis’
(Manning 2001:55).
In the following subsections, several forms of social deixis are focused on. All these
forms do not include the use of classifiers, even though some of the considered
languages do have inventories of classifying morphemes. In its turn, the ways of
expressing social relations with the use of classifiers are considered in Section 4. I
attempt to draw a parallel between different forms of social deixis which do not
involve the use of classifiers (Section 2.2), and the forms which do (Section 4).
As the basic definition of social deixis, I adopt the one proposed by Levinson
(1979:206) cited above and repeated in the following. Social deixis is:
those aspects of language structure that are anchored to the social identities of participants
(including bystanders) in the speech event, or to relations between them, or to relations
between them and other referents.

2.2. Examples of social deixis. Case studies


2.2.1. European languages: T/V pronoun phenomenon
The best known example of social deixis is the T/V phenomenon, which is almost
universal in European languages but is also attested elsewhere in the world (Foley
1997:314). The abbreviation T/V stands for a T form, named after Latin tu, second-
person singular pronoun, and a V form, from Latin vos, second-person plural. The first
form is typical for informal, relaxed usage, while the latter is for formal polite
contexts. Brown and Gilman (1972 cited in Foley 1997:314), who first discussed the
T/V phenomenon, described the use of the T/V pronouns along two dimensions:
power and solidarity. The dimension of power encodes power over another person
and the ability to control or influence his or her behaviour. The solidarity dimension
stands for the degree of closeness between interlocutors, presence of common

4
interests, and an assertion of at least some shared lived history. Figure 1 presents a
model of modern usage of T and V forms1.

Superior and solidary T V Superior and not solidary


(V) (V)
Equal and solidary Equal and not solidary
T V

(T) Inferior and solidary T V Inferior and not solidary (T)

Figure 1. Use of the T/V pronouns along two dimensions: power and solidarity (adapted from Foley
1997:316 citing Brown and Gilman 1972)

The power dimension encodes inequality in economic and political power, and the
different degrees of status this inequality entitles one to. As Foley (1997:314)
mentions, it is “largely linked to class or caste positions.” “Solidarity is closely related
to social roles, and their relative status entitlements, that the two interactants find
themselves in” (Foley 1997:315). If the interactants are equivalent in status, it is likely
that they assert solidarity between themselves. The vertical line in Figure 1 represents
the power dimension, the horizontal one – the solidarity dimension. Foley (1997:315)
draws attention to the fact that “the upper left and lower right boxes represent the
problematic cases, in which the power and solidarity dimension conflict: superior in
power, suggesting V, but solidary, claiming T, and inferior in power, marked by T,
but not solidary, indicated by V.” In modern usage the solidarity dimension has more
weight, in the first instance mutual T is required and mutual V in the second. Since
solidarity has become the primary dimension, the use of T forms has progressively
expanded in European societies (Foley 1997:316), though the parameters which
determine its use do vary there. English constitutes an exception, since it is a
European language which lacks a T/V contrast. It has other linguistic means to
indicate the contrast though, namely, usage of titles and names (which other
languages have additionally to the T/V distinction).
The process of changing one pronominal usage pattern to another, so called
“breakthrough”, was addressed in the study by Friedrich (1966) on pronominal usage
in 19th century Russian novels. Breakthrough can be exemplified, for instance, by the
switch to mutual usage of T forms between two individuals involved, which can result
in a move toward greater intimacy. It can also be illustrated by an opposite situation:
for instance the one described in Fillmore (1975:78, referring to Friedrich 1966),
where “two officers exchanging T while having a drink together will suddenly switch
to V when one of them feels insulted and issues a challenge to a duel. Thus, such
change in the forms is an indication that the speaker wants to reject previous

1 The usage of T and V forms until roughly 1800 was realised slightly differently. With the Industrial
Revolution of about 200 years ago, social stratification became much less rigid and increased
possibilities of mobility between classes began to emerge. As consequence, the dimension of solidarity
gained importance so that mutual T and V forms can be used between superiors and inferiors
depending on the degree of solidarity shared (Foley 1997:315)

5
assumptions of solidarity with his conversation partner. Likewise, it has been
observed that in Bolivian Lowland Spanish, parents address a child with usted (V
form) when they are angry with him/her, while domestic pets are always addressed
with V in that area (Crevels p.c.).
These examples show that the usage and interpretation of the personal pronouns in
question is highly dependent on discourse.

2.2.2. Sinhalese: pronoun system


The pronoun system in Sinhalese, a language of the Indic branch of the Indo-
European language family spoken on the island of Sri Lanka (Tilakaratne 1988:174),
provides another fine example of social deixis. The system encodes principles of
social relationship among speech act participants, thus showing a close
interrelationship between social stratification and language structure.
The pronominal system in spoken Sinhalese is complicated by variations
based on the social relationship between a speaker and his/her addressee. The 1st
person pronouns in this language do not vary according to social variables, but the 2nd
and the 3rd person pronouns do. We consider the 2nd person pronouns briefly below.

Masculine Feminine
Sing. Pl. Sing. Pl.
1. Formal written oba obala oba obala
2. Spoken: socially equal oya oyala oya oyala
3. Spoken: socially inferior umba/tho umbala/thopi umba/thi umbala/thopi
4. Spoken: socially upper obatuma obatumanla obatumi obatumila

Table 1. The 2nd person pronoun and social variables in Sinhalese (adapted from Tilakaratne
1988:179)

As can be seen from Table 1, the second person pronoun may have at least 4 different
registers. Line (1) presents the forms that are used in formal written speech and also
the spoken forms used by educated people. The other three registers are all
characteristic of a spoken Sinhalese; they differ with respect of relative social status
of speech act participants. Thus, when the speaker and addressee both belong to the
high or middle classes, they use the pronoun forms defined in line (2). When the
addressee belongs to the low social class, or when both the speaker and the addressee
do, the pronouns given in line (3) are employed. The middle social class or the upper
social class may not use these pronouns among themselves. Thus, when people climb
up the social ladder from low to middle class, they usually cease using these forms.
These pronouns are not affected by variation in age or gender. Interestingly, they all
have equivalents tho/thopi, thi/thopi with the difference that these forms are used
“when the speaker is in an angry mood or not well disposed towards the addressee”
(Tilakaratne 1988:179). Line (4) demonstrates the pronouns used by a speaker of the
lower class addressing a person of the upper social class, regardless of the formality
of the situation. These forms can be heard when the addressee is a political leader or a

6
high government official.
Unlike most European languages, Sinhalese shows a three-way distinction in the 3rd
person pronouns based on proximity between the speaker and the referent. Three
categories are distinguished: the first, proximal, has a meaning similar to that of the
English demonstrative pronoun ‘this’. The second, distal, is used when the referent is
a bystander and is similar to the English demonstrative pronoun ‘that’. The third
category pronoun is employed when the referent is in the vicinity of the speaker or at
another locale. Its English equivalent is third person pronoun ‘he’. All three categories
of the 3rd person pronoun have a similar stratification as the 2nd person pronouns
described above.
It should also be noted that verb forms vary as well along the social status
dimension in accordance with the variations of personal pronouns. An utterance will
be socially unacceptable if a verb form does not “socially fit” (Tilakaratne 1988:187)
the pronouns.

2.2.3. Guugu Yimidhirr: avoidance behaviour


Systems encoding social identities can be qualitatively different from the pronoun
systems considered above. Now we turn to the case of Guugu Yimidhirr, an
Australian Aboriginal language spoken in the area near Cooktown, Northern
Queensland, Australia. The Guugu Yimidhirr speaking society represents one of the
age-set formations where kinship links and age rank are the basic building blocks of
social structure. There the elder age grades usually hold the power and the highest
status is entitled to them. The case of Guugu Yimidhirr illustrates how exclusively
such variables as age, gender and kin relation can serve as the basis for differential
linguistic and non-linguistic behaviour. Section 4.2 in the following, considers the
influence of the kin and age parameters on the usage of classifiers in Jacaltec, Mam
and Akateko.
Characteristic of traditional Guugu Yimidhirr society is that social relations are based
entirely on kin. Guugu Yimidhirr speakers apply classificatory principles to extend
kinship to everyone they come into contact with. For instance, a term usually reserved
for aunts is also appropriate for an unrelated woman of the same great group (so-
called ‘moiety’) as the speaker and of the appropriate age. In the traditional Guugu
Yimidhirr society, certain relatives of a man were sacred, a ‘taboo’, to him. Among
these relatives was his wife’s mother, whose presence he had to avoid strictly. He also
avoided, but nevertheless could still have dealings with his wife’s father, and his
wife’s brothers, though in interaction with them he was obliged to use the special
“brother-in-law” language as a sign of respect. Transfer of information between a man
and his taboo relatives was indirect and mediated: a man directed his messages to his
in-laws via his wife using the brother-in-law vocabulary, and, in return, the wife’s
father communicated with his son-in-law via his daughter. A man could speak freely
with potential wives, for instance, with his wives’ sisters but he was obliged to speak
circumspectly with his own sister, who could never be his sexual partner. Haviland
(1979:215) mentions that “women were less likely to use special vocabulary with
their fathers- and mothers-in-law than were men, possibly because as members of

7
their husband’s group after marriage, they were in somewhat closer contact with their
in-laws than were men with theirs.” The opposite extreme was that with certain kins
one could enjoy extraordinary freedom. A grandparent of the same moiety, and of the
same sex, was a “prototypical joking partner, whom one could tease, insult, and goad,
in both word and gesture” (Haviland 1979:216). Thus, behaviour between grandfather
and grandchild of the same moiety, could go beyond the bounds of appropriateness.
Avoidance behaviour, though, was not the only way to express deference and
respect. If one had to speak, one had to mind how one spoke. To speak respectfully
meant to avoid the strident tones and rapid speech which was typical for ordinary
conversations. As Haviland (1979:217,223) puts it: “one speaks to a brother-in-law or
a father-in-law in a deliberately subdued voice, drawing out words and dropping into
a near whisper” because “speaking loudly and rapidly was associated not only with
familiarity and informality but also with anger and scolding”. Besides, one should
speak “sideways” trying not to face one’s interlocutor or address him/her directly.
Physical proximity with one’s in-laws was impolite. Haviland (1979:217) mentions
that in some areas of the Guugu Yimidhirr speaking area, it was said that a man would
speak to his father-in-law by addressing his dog instead.
As noted, traditional Guugu Yimidhirr had a special “brother-in-law” language
which was used when addressing the wife’s in-laws in a respectful manner. The
brother-in-law vocabulary was not in itself a full language separate from the ordinary
one. Instead it consisted of a rather small set of special words thought to be
appropriate for highly polite and respectful speech. Some everyday words might have
a one-to-one correspondence in the brother-in-law language, for instance, ‘echidna’:
balin-ga (common) vs. nhalngarr (formal). More often, though, several different
everyday words could be rendered by a single word in the brother-in-law language.
Consider the following example where the noun balin-ga ‘echidna’ and the verb
dhadaa ‘go’ are characteristic of the everyday language:
(1) Guugu Yimidhirr (Australian) (Haviland 1979:218)
a. balin-ga dhaday
‘Echidna went away’
In formal brother-in-law speech the same meaning can be expressed as follows
(Haviland 1979:218):
b. nhalngarr bali
‘Echidna went away’
Balil ‘go’ is also the brother-in-law equivalent of some other everyday verbs, with
such meaning as ‘crawl’, ‘paddle (in a boat)’, ‘float, sail, drift’, ‘limp’, ‘wade’, etc.
None of these everyday words could be used with a brother-in-law.
Some words, if they are pronounced in the proper slow and respectful manner, did
not require replacement when talking to one’s brother-in-law.
When the speaker had to treat the interlocutor with deference and respect, he partially
could do so by other available linguistic devices. He might use circumlocution or
stylized indirection by using a kin term in order to refer to the hearer without making

8
a direct reference, for instance as the English “allow me to escort Madam to the door”
(Haviland 1979:221). Another way is formally akin to the Guugu Yimidhirr brother-
in-law use of second-person plural pronoun, V form, considered in section 2.2.1. As
Haviland (1979:221) points out, development of the same device for showing respect
in such widely separated regions, “attests both to the universality of the problem –
expressing deference in speech - and to the naturalness of the solution”. This might be
because, as he speculates, a plural form literally makes more of the hearer, which is
more deferential.
Summarizing, the range of conventional social relations in Guugu Yimidhirr is rather
complex. There is strict avoidance between a man and his mother-in-law, an obscene
joking relationship between grandparent and grandchild, relaxed amicability between
friends and some family members, and there are various other forms of politeness and
restraint taking place between these extremes. Such a highly structured social universe
implies a set of speech styles or registers which are appropriate for different sorts of
interaction: informal, restrained and polite, differential and respectful, or deliberately
obscene. These registers comprise different sets of words and even non-linguistic
behaviour. A speaker of Guugu Yimidhirr can manipulate registers for a particular
purpose. Using a brother-in-law word where an everyday word would do, a speaker
conveys deference, or deliberately violating normal rules of speech politeness he can
insult or undermine a relationship. In more or less codified ways, the speech is partly
determined by the relationship between speakers and, in turn, determines the
relationship.
If we had to draw a parallel with the employment of T/V forms in some Indo-
European languages, the avoidance behaviour towards one’s in-laws as a sign of
deference and respect could be compared with the use of V form which also encodes
this meaning. The major difference is that the use of V form is mainly motivated by a
higher social position of one’s interlocutor or by his/her greater age. Avoidance
behaviour, on the other hand, is a means of showing deference which has kinship
motivations.

2.2.4. Japanese: honorifics


This subsection deals with another system of encoding social relations. As known,
numerous Asian languages have an elaborate system of honorifics, “grammatical
morphemes and special classes of words indicating social deixis among the
interlocutors or the referent of a participant in the utterance” (Foley 1997:319). By the
appropriate use of honorifics, one is able to label a referent or to identify oneself with
a certain social standing. Levinson (1979:207) suggested restricting the term
‘honorific’ to the cases in which the relations between speaker-referent, speaker-
addressee, speaker-bystander express relative rank or respect. This phenomenon will
be exemplified here by Japanese, since it is a language distinguished by a rich system
of honorifics (Inoue 1979; Shibatani 1990, cited in Foley 1997:319; Alpatov 2006).
First of all, it is necessary to mention that Japan has undergone drastic social and
institutional changes in the period from the summer of 1945 until the early fifties.

9
Nowadays, Japan is unmistakably a country of middle-class citizens: an annual
government poll have consistently indicated since 1969, that ninety percent of the
Japanese people consider themselves to belong to the middle class (Inoue 1979:290).
But as any other society, regardless of the form of the government, Japanese society
has social stratification. The language reflects this situation most closely, and the
norms of appropriate linguistic behaviour are based on the way in which the society is
stratified. Some factors determining the norms are quite universal, some are uniquely
Japanese.
Honorifics in Japanese fall into two categories. Terms used to describe them vary
slightly in different sources. To name just a few: Alpatov (2006:4) refers to them as
addressive category and honorific category. Foley (1997:319, citing Levinson
1983:96), labels them addressee honorifics and referent honorifics. In the present
description, I keep to the latter terms, since they seem more precise to me.
Addressee honorifics are used to indicate deference by the speaker to the
addressee. The category of addressee honorifics distinguishes two grammatical forms:
the form indicating deference to an addressee who is a stranger or a person of a higher
status with whom solidarity cannot be claimed, and a neutral form asserting an
equivalent, solidarity relationship to the addressee. Foley (1997:319) signals the
equivalence of the first and latter forms with the T/V phenomenon. The following
examples demonstrate polite (2a) and neutral (2b) forms:
(2) Japanese (Japanese-Korean) (Shibatani 1990:377, cited in Foley 1997:319):
a. watakushi kare ni ai-mas-u
I he DAT meet-POL-NPST
‘I'
ll see him’
b. boku kare ni au yo
I he DAT meet PRTCL
‘I'
ll see him’

In referent honorifics, deference is accorded by the speaker to the referent of a


nominal participant in his utterance (Foley 1997:319, Alpatov 2006:38). Foley
(1997:319) points out that the European languages have no equivalent strategy.
Referent honorifics can be realized in three grammatical forms: honorific forms,
humbling forms, and neutral or non-deferential forms. The choice of an appropriate
grammatical form of either of the categories is determined by a range of factors which
can be placed on two dimensions.
On one dimension, there are the notions superior - equal - inferior. Whom the
speaker interprets as superior, inferior or equal depends on several factors. First, the
social position of an interlocutor plays a role. A speaker’s assignment of an
interlocutor of a particular social position is often very subjective. Second, age grade
determines who is perceived as superior, inferior or equal. People older than a speaker
are considered superior, younger than the speaker are often taken as inferior. Third,
gender of an addressee or referent sometimes still plays a role, where a man is
considered higher on the social hierarchy. Besides these relational factors, situational

10
ones are also at work. Hatsutaro (1968:86-104, cited in Alpatov 2006:18) and Inoue
(1979:291) point out that asking a favour creates a situation of inequality between
interlocutors. It can be exemplified by the customer-salesperson relationship where
the former has decisive power over the latter. “While the customer has a choice of
three styles of speaking, from a very courteous style to a rather rude one, a well-
trained salesperson must remain courteous regardless of the customer” (Inoue
1979:291). Alpatov (2006:18) mentions that during the war years, consumption was
not interpreted as a favour due to the shortage of goods; as a consequence, the speech
of salespeople was reported to be less polite.
On the second dimension, there is the distinction between ingroup vs.
outgroup. The notion of ingroup can extend beyond one’s immediate family. Thus, it
contrasts members of a family and organizations perceived in terms of a family, the
ingroup, with people outside the family or organization. Besides, who will be seen as
a member of the ingroup, depends largely on the presence of other people in the
interaction. Among ingroup members, one uses honorifics in reference to an insider to
signal his superiority and power. In the presence of a person who is not ‘a member’,
the use of honorifics referring to the insider is not appropriate. In such case, a speaker
should present the whole ingroup as a unit of equals though inferior to an addressee.
Quite unacceptable is the use of honorifics referring to oneself, independently of the
social status one may have. When talking about members of the outgroup, honorifics
are necessary in order to reflect their social distance. Consider the example of brothers
speaking with each other about their father: the brothers will use honorifics to reflect
their father’s seniority in age and power. However, when speaking about their father
in presence of, or to someone who is not a member, the father will be considered as a
member of ingroup, as equal to oneself, and no honorific forms must be employed
(Alpatov 2006:19), (Inoue 1979:293). A similar situation occurs within a corporation.
Colleagues would use honorific forms in referring to the company president when
talking among themselves (ingroup), but when speaking to an outsider (outgroup), no
honorifics must be used. The difference is illustrated in examples (3), where (3a)
shows reference to the company president between the ingroup members and (3b)
shows how the same company president would be referred to when speaking with an
outsider:
(3) Japanese (Japanese-Korean) (Shibatani 1990:379, cited in Foley 1997:322)
a. shatyoo san wa ima o-dekake ni nat-te i-mas-u
president Mr TOP now HON-go DAT become-ing go-POL-NPST
‘The president has gone out now’
b. shatyoo wa ima dekake-te ori-mas-u
president TOP now go out-ing be.HUM-POL-NPST
‘The president has gone out now’
This behaviour pattern though is not universal in Japanese society, since not all
engage in occupations causing them to regard themselves as belonging to a family-
like ingroup, like a corporation. Inoue (1979:293) presents the examples that
university professors regard their students as members of their group but do not

11
consider the wider university to be part of such. Physicians who work at hospitals, are
not group oriented either.
In the process of conversation, speech register can shift from deferential and
respectful to neutral, and vice versa. The first situation is relevant when interlocutors
become more close to each other. The latter can be characteristic of disputes and
quarrels. Thus, a speaker can shift to exaggeratedly honorific speech if (s)he wishes to
create some distance with her/his addressee. Sudden change in register during an
interaction can carry an ironic connotation, and can be used to create a mocking effect
(Alpatov 2006).
Another point that must be mentioned is the difference between men’s and
women’s speech in Japanese. “Women have traditionally been trained to speak more
politely than men among themselves as well as towards men” (Inoue 1979:293).
Besides, the language has certain forms which are used only by men, or only by
women, thus illustrating the case of authorized speakers as defined by Fillmore (1975,
cited in Levinson 1979:207). As Alpatov (2006:138) notices, the men’s and women’s
speech has partially assimilated in the past few years: specific female words and
affixes can appear now in men’s speech and vice versa. Such linguistic changes are
undoubtedly connected with the changes in social structure.
As far as written speech is concerned, the use of forms of the addressive or
honorific categories is slightly different. When an addressee is presupposed (e.g.
letters), types of written speech have a higher level of politeness than oral speech.
Thus, if in oral speech one uses neutral forms to an addressee, in a letter, one sooner
uses forms expressing deference even when referring to the same person (Alpatov
2006:32). In other types of written speech, when a writer has no personal bond with
an addressee (such as scientific texts, newspaper articles, etc.), mainly neutral forms
are used. The use of honorifics in written speech is rather limited. The only case in
which honorifics are a norm in any type of text is in the one containing reference to
the emperor and his family. Japanese used to have a special vocabulary and special
grammatical forms which were employed exclusively in reference to the emperor. In
contemporary Japanese these forms are being avoided, and ‘usual’ honorific forms are
preferred (Alpatov 2006:43). It is interesting that neither in references (in texts) to
foreign monarchs, nor to other top (government) officials, honorific forms are
normally employed.
Summarizing, it is necessary to point out that, as Inoue (1979:295) mentions,
“deferential language in Japanese does not mark social class in any simplistic way for
either the people spoken to, or the people spoken about. Everyone in Japanese society
has the chance to use all of the deferential expressions in his daily life, depending on
the various circumstances in which he finds himself.” Interestingly, deferential
language may sooner be a marker of the social identity of a speaker. In modern times,
knowledge of deferential expressions is a linguistic prestige feature. Thus, the ability
to use such expressions appropriately is considered to be a mark of good education
and a good upbringing. Well-educated intellectuals are generally more conservative
regarding the use of deferential expressions. According to Inoue (1979:296), they tend
to speak in a more polite language regardless of who they are speaking to. This can be

12
seen rather as a reflection of their estimate of their own social position, than a
reflection of the respect of others. But what is considered as appropriate does change
as Japanese society as a whole changes and adjusts itself to the changes in the larger
world. Thus, honorific language conveys complex messages not only about
interpersonal relations between a speaker and other interlocutors in the society, but
also about the speaker himself.
Such a use of honorifics in function of social deixis can be compared to the
use of ‘honorific’ classifiers found in Pohnpeian (see Section 4.1).

2.2.5. Samoan: respect vocabulary


The given section considers the case of Samoan (Austronesian). This language is
characterized by the presence of a special ‘respect’ vocabulary, usually referred to as
upu fa' aaloalo ‘respectful words’, which are distinguished from more ordinary
‘common’ terms. These respectful words describe individuals and groups, some of
their actions, attributes and possessions. They are used to address or refer to people
with high social status, such as titled people, including chiefs and orators. For
instance, the term ‘come’ has three forms in Samoan (Duranti 1992:84, cited in
Merthrie 2000:202):
sau (common)
afio mai (used for chiefs)
maliu mai (used for orators)
Below is an example that illustrates the use of the term maliu mai in reference to a
senior orator:
(4) Samoan (Austronesian) (Duranti 1992:84, cited in Merthrie 2000:202)
ia '
o le maliu mai laa o le Makua
‘so the senior orator has arrived’
Nevertheless, the function of this respectful vocabulary is not only limited to this
usage. Duranti (1992, cited in Merthrie 2000:202) has found that respectful words
were sometimes used for people who were not entitled to them, whereas those who
were entitled to them did not always receive them. The usage of the words can be
rather inconsistent. Namely, in reference to the same individual different terms might
be employed: one respectful and the other one common. The choice between the two
can be used to highlight different aspects of the identity of the person addressed or
referred to. Thus, similarly to honorific terms in Japanese, these terms construct a
variable relationship between the speaker and the person being talked to or about,
with the ability to transfer attitude or by emphasizing one’s power and status or by
depriving the referent of such.

2.3. Summary
As Haviland (1979:209) puts it “language does not exist and utterances do not occur
in a social vacuum.” Social relations are reflected in all languages, but the languages
differ in the forms they use. The present section considered several case studies

13
demonstrating different forms of linguistic and non-linguistic behaviour, which reflect
or are determined by social identities of speech act participants, e.g. speaker and
addressee, speaker and referent, speaker and bystander, or speaker and setting. The
case studies show that social identities of the participants in a speech event can be
anchored to various aspects of language structure. Section 4, deals with four case
studies in order to see whether and how classifier morphemes can also function as
social deictic items. First, however, the phenomenon of classifiers is considered in
more detail.

3. Classifiers
3.1. Systems of nominal classification and the place that classifiers hold among
them
Classifiers represent only one of several types of nominal classification systems. The
specific interest in classifiers raises the question of the typology of nominal
classification, and therefore the present section offers a brief overview of nominal
classification devices categorized so far and the place that classifiers hold among
them.
The typology of nominal classification systems has been worked out in
Grinevald (2000), which stands now as the most comprehensive approach to the issue.
Based on morpho-syntactic criteria, Grinevald places types of nominal classification
systems along a continuum, running from purely lexical to purely grammatical means
of classification, allowing for overlap. Classifiers, the main interest of the present
paper, are placed at a mid-way point of the continuum.

<---------------------------------------------------------------------------------------------------->
Lexical Lexico-grammatical Grammatical
Measure terms, class terms Classifiers Noun classes, gender

Figure 2. Continuum of nominal classification systems (adapted from Grinevald 2000:61)

“ ‘Lexical’ here means (a) part of the lexicon and its word-building dynamics and (b)
semantically compositional” (Grinevald 2000:55). On the lexical end of the
continuum Grinevald places measure terms and class terms, which she clearly
differentiates: “Measure terms are lexical in the sense that they are semantically
compositional / analytic noun phrases, and class terms are lexical in the sense that
they operate like derivational or compounding morphology at word level” (Grinevald
2000:58). Measure terms exist in all the languages of the world and express
quantities. Examples of measure terms in English can be those of mass nouns and of
count nouns, respectively: a pound of sugar, a slice of bread and a line of cars, a pile
of books. As Grinevald mentions, these measure terms are sometimes labelled
classifiers, what can make any language to be a classifier language. The ‘true’
classifiers “categorize by some inherent characteristic of the referent, such as its shape

14
(and texture, or material, etc.)” in contrast with measure terms (Grinevald 2000:58),
for instance:

[Classifier] [Mensural term]


two [ROUND] oranges two [bags of] oranges
a [FLAT.FLEXIBLE] shirt a [stack of] shirts
three [HUMAN] children three [circles of] children

Class terms are defined by Grinevald (2000:59) as “classifying morphemes which


participate in the lexico-genesis of a language”. The equivalent of such class terms in
a language like English would be morphemes like -berry or -tree (as in strawberry,
raspberry; palm tree, apple tree), or ‘agent’ morphemes -ist (as in novelist, chemist), -
er (as in singer, baker), -man (as in police man, mail man). In some languages the
relation between class terms and classifiers is so close, that it leads to some confusion
of these two types of nominal classification. As a result, there are languages that have
been considered to have classifiers when in fact they have class terms. According to
Grinevald (2000:59), this has been the case for the Rama language (Chibchan,
Nicaragua; Craig 1990a,b), where some morphemes are quite productive. The relation
between class terms and classifiers is in fact very close. For instance, in the Thai
language, family classifiers originated as class terms, and both systems still co-exist
(DeLancey 1986, cited in Grinevald 2000:60).
The ‘Grammatical’ end of the continuum stands for “part of the morphosyntax
of a language” (Grinevald 2000:55). At this end, Grinevald places gender and noun
class systems, which are considered to belong to the same basic type (Aikhenvald
2000:19). As Grinevald (2000:57) adds, gender and noun classes are treated often as
one major system of nominal classification, which is contrasted with classifiers. Both
gender and noun class systems are in fact agreement systems. Gender is a type of
nominal classification, which is not always overt in the noun itself “but is echoed in
other elements of the clause which vary according to the language” (Grinevald
2000:55). Gender, by and large, is semantically empty, it is assigned to all nouns of
the language and it is obligatorily marked. Noun class systems are regarded as typical
of languages of the Niger-Congo linguistic stock, especially the Bantu family. Bantu
noun class systems are reported to have up to twenty morphological classes. Like the
above mentioned gender systems, they exhibit agreement patterns within the NP and
across to the predicate. Consider the Bantu language Tonga in example (5):
(5) Tonga (Niger-Congo) (Collins 1962, cited in Allan 1977:286)
ba-sika ba-ntu bo-bile
CL:PL.HUM-have.arrived CL:PL.HUM-man CL:PL.HUM-two
‘Two men have arrived’

An intermediate, lexico-grammatical, position on the continuum of nominal


classification is occupied by classifiers. This can be explained by the following
reasons. First, as Grinevald (2000:61) puts it, they are distinct from noun class/gender

15
systems [grammatical end] in their incomplete grammaticalization. This is evidenced
by general properties of classifiers (Dixon 1982, 1986, cited in Grinevald 2000:62):
their ability to add semantic content to the noun phrase in which they occur, their non-
obligatorily use, their sensitivity to discourse, variation in register, possibility to
assign a noun to various classes at speaker’s will, as well as the fact that they
constitute an open system and assign nouns into a largish number of classes. Second,
classifiers are different from measure terms/class terms [the lexical end of the
continuum] in that they mark a category of a noun “beyond the noun word itself, in
independent morphemes or in affixes on other elements of the clause” (Grinevald
2000:61).
The intermediate position of classifiers on the continuum reflects also the
possibility for overlap/blending between the systems of nominal classification, which
results in the borrowing of characteristic features of either the lexical or grammatical
system, thus creating using Grinevald’s (2000:81) terminology, “the fuzzy edges”.
Thus, classifiers are only one of several types of nominal classification
systems. In the present thesis I focus exclusively on this type. The following
subsections consider different types of classifiers.

3.2. Types of classifiers.


Summarizing the criteria discussed in Dixon (1982, 1986 cited in Grinevald 2000:61)
I define classifiers here as:
Morphemes which classify and categorize nominal referents according to their specific
characteristics, and which tend to have the following properties:
• ability to add semantic content to the noun phrase in which they occur;
• non-obligatorily use;
• sensitivity to discourse, variation in register;
• possibility to assign a noun to various classes at speaker’s will;
• constitute an open system.
Classifiers as a type of nominal classification system are usually divided into several
sub-types: numeral, noun, genitive, verbal, deictic and locative classifiers, where the
latter two are “still in need of justification” (Grinevald 2000:623). The classifiers owe
their names to the morphosyntactic environment where they occur, or, in other words,
classifiers are labelled according to the morpheme they are closest to or attached to.
All the mentioned sub-types will be illustrated and exemplified next.

3.2.1. Numeral classifiers


In some languages, when counting inanimate as well as animate referents, the
numerals will concatenate with certain morphemes, the so-called numeral classifiers.
Thus, numeral classifiers are morphemes that appear contiguous to numerals or to
quantifiers. Allan (1977:286) labels this type as ‘the paradigm type’. In some numeral
classifier languages not necessarily every noun can be associated with a classifier.
There are nouns that take no classifier and nouns which can take various classifiers
depending on the property of the referent noun one wishes to accentuate.

16
Numeral classifiers characterize their referents according to semantic criteria. They
can provide a division of nouns into human and non-human, or animate and
inanimate. It can be a three-way distinction when nouns are divided into humans, non-
human animates and inanimates. The examples (6) and (7) come, respectively, from
Pohnpeian, a Micronesian language, and the Yucuna language of South America,
where the numeral classifier refers to animacy of the referent noun:
(6) Pohnpeian (Micronesian) (Rehg 1981:130, cited in Grinevald 2000:63)
pwihk riemen
pig 2+CLF:AN
‘two pigs’
(7) Yucuna (Arawakan) (Aikhenvald 2000:106)
pajluhua-na yahui
one-CLF:AN dog
‘one dog’
Classifiers used with human referents can encode a complex meaning, including
gender of the referent and attitude of the speaker to the referent. This is the case for
Assamese, where four of the ten classifiers are employed with humans and have the
following semantics:
z n: human males of normal rank (respectful);
g raki: humans of either sex (respectful);
z ni: female animals; human females (disrespectful);
z na: high-status humans of any sex;
The use of the classifiers is illustrated in example (8).
(8) Assamese (Indo-Aryan, Indo-European) (Barz and Diller 1985:169, cited in
Aikhenvald 2000:102)
a. tini z  n x  khi
three CLF:HUM.M friend
‘three friends’ (respectful)
b. tini z  ni sowali
three CLF:HUM.F.DISRESP girl
‘three girls’ (disrespectful)
c. tini g raki m  hila
three CLF:HUM.RESP woman
‘three women’ (respectful)
Inanimate referents can be classified according to their physical properties (such as
shape, consistency, size, boundedness, etc.), as can be seen from the following
examples.
(9) Japanese (Japanese-Korean) (Matsumoto 1993, cited in Grinevald 2000:63)
enpitsu ni-hon
pencil two-CLF:LONGISH

17
‘two pencils’
(10) Ponapean (Micronesian) (Rehg 1981:130, cited in Grinevald 2000:63)
kehp rioumw
yam two+CLF:BAKED
‘two yams’
Concerning morphological realization, a numeral classifier can be affixed to a
numeral or realized as an independent lexeme. As can be seen from the example (9), it
can be realized as a suffix, or more rarely, as a prefix, for instance as in Itonama, an
isolate spoken in the Bolivian Amazonian lowlands (Crevels, p.c.). Some languages
demonstrate a more unique pattern, such as realization of numeral classifiers as
infixes. This is the case, for instance, for Baure and Yagua, where classifiers are
infixed to numerals, which are usually small numbers:
(11) Yagua (Peba-Yaguan) (Aikhenvald 2000:107)
tá-dasiy-quii
one-CLF:THIN.POLE-one
‘one (shotgun, blowgun, palm trunk, etc.)’
As noted, numeral classifiers can also take form of independent lexemes. Aikhenvald
(2000:104) points out that this type has mostly been encountered in isolating
languages. We can turn to Thai to illustrate the case:
(12) Thai (Tai) (Allan 1977:286)
khru lâ·j khon
teacher three CLF:PERSON
‘three teachers’
As far as areal distribution goes, numeral classifiers are present in most languages of
mainland Southeast Asia (Thai, Burmese, etc.) and in many other East Asian
languages, such as Japanese and Chinese (Grinevald 2000:63). They are encountered
in most Western Austronesian and Oceanic languages, though absent from the
Austronesian languages spoken in Taiwan (Aikhenvald 2000:124). A whole range of
South-American Indian languages have large sets of numeral classifiers, as well as
numerous Mexican and Central American languages. They are scattered across North
America (Sherzer 1976; Cambell 1997, cited in Aikhenvald 2000:121). Numeral
classifiers are hardly found in the languages of Africa, Australia or Eurasia.

3.2.2. Noun classifiers


Grinevald (2000:64) characterizes noun classifiers as “free morphemes standing in a
noun phrase, next to the noun itself or within the boundaries of the noun phrase”.
Additionally she mentions, that “they are crucially found independently of the
operation of quantification”.
Aikhenvald (2000:81) adds another property of noun classifiers to this
definition: their scope is a noun phrase and not a noun by itself. They are “a type of
non-agreeing noun categorization device”, “their choice being determined by lexical

18
selection.”
Noun classifiers are distributed less widely in the world then numeral
classifiers, but they are found in numerous Australian languages. Here the languages
spoken along the east coast of the Caper York peninsula of Australia must be pointed
out (Aikhenvald 2000:97). Further, they are present in a few Western Austronesian
(Marnita 1996; Durie 1985, cited in Aikhenvald 2000:97), and Oceanic languages
(Jensen 1977, cited in Aikhenvald 2000:97). Noun classifiers are found in some
Mesoamerican (such as the Mayan languages of the Kanjobal branch) and might be
present in a few South-American languages, however it should be investigated
further.
The semantics of noun classifiers relate to the inherent nature of the object and
to its functions. As Aikhenvald (2000:283) states, “there is usually a classifier for
humans, or animates, one or more classifier(s) for non-human animates, and a number
of classifiers for inanimates based on their inherent properties or functions.” Noun
classifiers can be exemplified by the following sentences taken from Jacaltec, where
the referent nouns are categorized as belonging to a class of male humans (naj),
 
animals (no ) or plants (te ).
(13) Jacaltec (Mayan) (Craig 1986:264, cited in Grinevald 2000:65)
a. xil naj xuwan no  lab’a
saw CLF:M.NKIN John CLF:ANIMAL snake
‘(man) John saw the (animal) snake’
b. kaj te  tahnaj ixpij
red CLF:PLANT ripe tomato
‘The ripe (plant) tomato is red’
The following example comes from Dâw, an indigenous language of South America,
where any noun with generic reference can be used as a noun classifier, e.g. dâw
‘human’:
(14) Dâw (Makuan) (Martins and Martins 1999:258)
dâw-tog
CLF:HUM-girl
‘a girl’
The semantics of classifiers used for categorization of human beings differs from
language to language. As this is the major interest of the present paper, it will be
considered in more detail in Section 4. Running ahead, it can be mentioned that
humans further can be categorized by noun classifiers according to such parameters as
age, sex, kinship relationship, social status, respect, etc.

3.2.3. Genitive classifiers


Classifiers used in possessive constructions are referred to as genitive classifiers.
Under this cover term Grinevald (2000:66) subsumes classifiers that are labelled
“possessed, possessor, possessive, relational” by other authors. Aikhenvald
(2000:125) divides ways of categorizing nouns in possessive constructions into three

19
categories:
• when the classifier categorizes the possessed noun; she labels this type
“possessed classifiers”. In this case the choice of the classifier can be
determined by the nature of the referent of the possessed noun in terms of its
animacy, physical properties (such as shape), function and other.
• when the classifier categorizes the possessor, we have the so-called “possessor
classifiers”. In this construction the choice of the classifier is conditioned by
the properties of possessor. Aikhenvald (2000:139) writes that this type is
extremely rare.
• when classifiers characterize the relationship between the possessor and the
possessed depending on how the possessed object can be handled or used, we
have the so-called “relational classifiers”.
Each of these categories of genitive classifiers will be briefly considered next.
Possessed classifiers tend to characterize the possessed noun in terms of its animacy
or/and physical properties (shape, size, consistency); they can also categorize the
referent noun in terms of its function. The following example is taken from Panare,
where classifying morphemes correspond to generic nouns, categorizing food,
animals, weapons, vehicles and manufactured objects:
(15) Panare (Cariban) (Mattéi-Müller 1974, cited in Grinevald 2000:66)
y-uku-n wanë
1SG-CLF:LIQUID-GEN honey
‘my honey (mixed with water for drinking)’
In Palikur referents are classified depending on their function, or the ways in which
they can be handled: fruit can be eaten or planted, animals can be domesticated or
caught for food. The possessed classifier is also in a generic-specific relationship with
the noun it refers to. Thus, in Palikur (Arawakan), the form pig ‘pet’ is used with
domesticated animals and win ‘catch’ is used with animals that are caught to eat:
(16) Palikur (Arawakan) (Aikhenvald 2000:142)
a. gi-pig pewru
3M-pet dog
‘his dog’
b. ni-win arudiki
1SG-catch tapir
‘my catch-tapir (the tapir I caught)’
According to Aikhenvald (2000a:147), possessed classifiers are found in a number of
languages of North and South America. They are also encountered in some multiple
classifier languages spoken in Northern China and Indochina, and in Papuan
languages of Central and Southern Bougainville.
Possessor classifiers are, according to Aikhenvald (2000a:139), extremely rare in the
languages in the world. They are reported in only two languages of South America.
Available examples demonstrate its choice is based on animacy distinctions of the

20
referent noun. However, as Aikhenvald (2000a:293) mentions, no reliable
generalizations on the semantics of possessor classifiers can be made, since the
number of examples is very limited.
Possessor classifiers are so far found in Dâw and, possibly, in Hupda (Moore
and Franklin 1979, cited in Aikhenvald 2000:139), of the Makuan family spoken on
the Brazil-Colombia border. The following examples come from Dâw, where
constructions with alienably possessed nouns require possessor classifiers:
(17) Dâw (Maku) (Martins 1994:138-41, cited in Aikhenvald 2000:139)
a. yud dâw-tôg- j
clothing human-female.child-CLF:AN.POSSR
‘girl’s clothing’
b. yak kaw-wâ´-d :´
manioc garden-up-CLF:INAN.POSSR
‘manioc from a garden’

Relational classifiers tend to divide nouns into consumable and non-consumable.


Further, objects can be classified according to the ways in which they can be
consumed or prepared, or on how they were acquired (e.g. found, or received as a
gift).
An example of relational classifiers can be found in Kipeá. This language has
a system of 12 classifiers characterizing the relationship between the possessor and
the possessed in terms of how the possessed could be handled by the possessor. In the
following examples food is classified depending on how it was prepared, by boiling or
roasting, or how it was acquired:
(18) Kipeá (Extinct, Macro-Ge) (Rodrigues 1995, cited in Aikhenvald 2000a:135)
a. dz-upodó do
sabuka
1SG-CLF:ROASTED POSS fowl
‘my fowl (roasted)’
b. dz-udé do sabuka
1SG-CLF:BOILED POSS fowl
‘my fowl (boiled)’
c. dz-ukisí do sabuka
1SG-CLF:SHARING POSS fowl
‘my fowl (that was my share)’
d. dz-ubá do sabuka
1SG-CLF:GIFT.FROM.OUTSIDERS POSS fowl
‘my fowl (that was given to me)’
A parameter such as value is more frequently encoded by this classifier type than by
the other ones (Aikhenvald 2000:133). Raga, among other Oceanic languages from
Vanuatu, has special classifiers for valued possessions:

21
(19) Raga (Austronesian) (Lichtenberk 1983:154, cited in Aikhenvald 2000:294)
qoe pila-ma
pig CLF:VALUED.POSSESSION-your
‘your (SG.) (valued) pig’
Relational classifiers can also reflect the closeness of the relationship between
possessor and possessed. This is the case for the Hohôdene dialect of Baniwa where
two relational classifiers correlate with the degree of closeness of the possessed to the
possessor. In the following example t inu ‘dog’ has an intimate relationship with the
possessor, and therefore the noun takes the suffix -ni:
(20) Hohôdene dialect of Baniwa (Arawakan) (Aikhenvald 2000:143)
a. nu-t inu-ni
1SG-dog-POSS
‘my dog (the one I brought up)’
In the next example, t inu ‘dog’ is not so close to the possessor, so it takes the suffix -
te:
b. nu-t inu-te
1SG-dog-POSS
‘my dog (the one I found)’
Some systems of relational classifiers can express a social function. For instance,
Pohnpeian has six relational classifiers referring to kinship. Depending on which
classifier is used, e.g. the head noun pwutak ‘boy’ can take the following meanings:
(21) Pohnpeian (Micronesian) (Rehg 1981:183, cited in Aikhenvald 2000:294)
a. pelie pwutak
CLF:PEER boy
‘his peer-group boy’
b. rie pwutak
CLF:SIBLING boy
‘his brother’
c. wahwah pwutak
CLF:NEPHEW boy
‘his nephew
d. kiseh pwutak
CLF:RELATIVE boy
‘his boy relative’
e. sawi pwutak
CLF:CLAN.MEMBER boy
‘his boy clansmember’
f. nah pwutak
CLF:DOMINANCE.OF.THE.POSSESSOR boy
‘his son’

22
The classifiers in Pohnpeian will be considered in more detail in Section 4.1.
Relational classifiers are present in Oceanic and Micronesian languages. They are also
found in small numbers in Austronesian languages spoken in Papua New Guinea.
Also two South American languages are reported to have inventories of relational
classifiers (Aikhenvald 2000:147).

3.2.4. Verbal classifiers


Allan (1977:287) labels the languages with this type of nominal classification as
“predicative classifier languages” due to the fact that verbs in some languages consist
of a stem which varies according to certain characteristics of the object “conceived as
participating in an event whether as actor or as goal”. As Grinevald (2000:67) puts it,
verbal classifiers do not “classify the verb itself but rather one of the nominal
arguments of the verb”. Depending on the transitivity of the main verb, verbal
classifiers can categorize the object or the subject of a sentence.
On the basis of morphological criteria verbal classifiers can be roughly divided
into two categories: (a) verbal classifiers that are found inside the verb form and (b)
verbal classifiers realized as affixes. The first category when a classifier is
incorporated into a verb, can be exemplified by Mayali, where, in this case, a verbal
classifier refers to a noun in the object function.
(22) Mayali (Australian) (Evans 1996, cited in Aikhenvald 2000:150)
ga-yaw-garrm-e al-daluk
3/3-CLF:BABY-have-NP CLII-woman
‘She has a baby girl’
Examples of verbal classifiers realized as affixes can be found in South American
languages. For instance, Terêna, a language spoken in Brazil, has several dozen verbal
classifiers which characterize the subject/object argument in terms of animacy and
physical properties (shape, size, form):
(23) Terêna (Arawakan) (Aikhenvald 2000:152)
oye-pu i-co-ti
cook-CLF:ROUND-THEME-PROGR
‘He is cooking (round things)’
Lengua Mascoy (Maskoian), a language of Paraguay, has verbal classifiers which
marks such notions as direction and location (Adelaar with Muysken 2004:498).
Itonama, an isolate spoken in the Bolivian Amazonian lowlands, employs
classifiers in existential constructions and with locational and positional verbs. The
choice of a classifier is based on animacy, shape and position of the referent. Example
(24) illustrates the use of the classifier -du which refers to oval objects + singular:
(24) Itonama (Isolate) (Crevels 2003)
nik’abï osi-du upala karomaya
there be-CLF:OVAL.SG stone black
‘There is a black stone over there’

23
Verbal classifiers are so far encountered in North and South America, in Australia, in
Papuan languages of southern and central Bougainville (Aikhenvald 2000:171).

3.2.5. Locative classifiers


Locative classifiers are “morphemes which occur in locative noun phrases”
(Aikhenvald 2000:172). According to Aikhenvald (2000:172), this type of classifier is
rare, and all available examples come from several South American Indian languages.
Thus, locative classifiers can be exemplified by Dâw, where the choice of a
classifier depends on physical properties of the referent, e.g.: k d ‘inside a bounded
object’, m ’ ‘inside liquid, or fire’, bit ‘underneath an object with an upper boundary’,

w ‘above an unbounded object’, a ‘inside a mixture’.
(25) Dâw (Maku) (Aikhenvald 2000:174)
a. xoo-k d
canoe-IN:HOLLOW
‘in a canoe’
b. nââ-pis-m
water-small-IN:LIQUID
‘in a small river’
In all examples found so far, locative classifiers are fused with an adposition.

3.2.6. Deictic classifiers


Both locative and deictic classifiers are described by Grinevald (2000:68) as “minor
types” and “in need of justification”.
Deictic classifiers “obligatorily occur with deictic elements such as
demonstratives and articles” and “they characterize the noun in terms of its shape,
animacy and position in space and they do not always appear on the noun itself”
(Aikhenvald 2000:176).
This type of classifier can be exemplified by Itonama. This language is
reported to have 17 demonstrative classifiers2, the choice of which depends on
number (singular or plural), animacy, position and shape of a referent object (Crevels
2003). In the following example the classifier –ba categorizes the referent object as
‘long, winding’:
(26) Itonama (Isolate) (Crevels 2003)
k-a’-ki-tya-ne no’o-ba makaya kahana’-na
F-2SG-IMP-wash-NEUT DEM.PROX-CLF:LONG.WINDING clothes old-NEUT
‘Wash these old clothes!’
Deictic classifiers are encountered in some languages of North and South America,
and in Eskimo (Aikhenvald 2000:177).

2 The same set of classifiers is used also on verbs (Crevels, p.c.)

24
3.3. General remarks on classifiers
Classifiers have semantic and discourse-pragmatic functions. Classifiers provide
information about classes of units. Since they carry different information from that of
the noun, classifiers can add information to it. This is salient in cases when a referent
noun is not associated with one single classifier and can be assigned by a speaker to
different categories, depending on what property of the referent s/he wishes to
emphasize. For instance, in Trukese (Austronesian), the analysis of the number of
classifiers a noun can occur with and the change in meaning that a change of a
numeral classifier entails, show that out of 821 nouns in the data, 5% are uncountable,
67% are normally associated with only one classifier, 12% can occur with more than
one classifier with no change of meaning, and 16% occur with more than one
classifier but with a change in meaning (Benton 1968:110-111,139, cited in Adams
1986:242). Thus, ability to add semantic context to the noun phrase constitutes an
important property of classifiers. This can be nicely exemplified by Burmese (Tibeto-

Burman), where one and the same referent noun myi ‘river’ can have at least eight
different readings, depending on which classifier it is used with (Becker 1975:113):
myi  t ya  ‘river one place’ (e.g. destination for a picnic)
myi  t tan ‘river one line’ (e.g. on a map)
myi  t hmwa ‘river one section’ (e.g. a fishing area)
myi  t ‘sin ‘river one distant arc’ (e.g. a path to the sea)
myi  t w ‘river one connection’ (e.g. tying two villages)
myi  t ‘pa ‘river one sacred object’ (e.g. in mythology)
myi  t khu’ ‘river one conceptual unit’ (e.g. in a discussion of rivers in general)
myi  t myi  ‘river one river’ (the unmarked case)

It is difficult to say, though, whether such types of variation in the use of classifiers
occur in all classifier languages. As Adams (1986:244) points out, “it is probable that
the larger the set of classifiers in a language, the more it will enhance the number of
choices.”
Beside the above noted semantic function, classifiers can have discourse-
pragmatic functions. According to Aikhenvald (2000:333), the use of classifiers
which are not obligatory correlates with “referentiality, specificity, definiteness,
topical continuity, and the salience in discourse” of the referent noun. Obligatory
classifiers have a similar tendency though display it to a lesser degree. As she further
mentions, all classifiers can be used anaphorically. Anaphoric use of classifiers can be
exemplified by Kilivila:
(27) Kilivila (Austronesian) (Senft 1996:21, cited in Aikhenvald 2000:331)
a. a-tatai tataba
1SG-carve tataba_board
‘I carve a tataba-board’
b. tauwau tabalu m-to-si-na
men Tabalu_subclan this-CLF:M-PL-this

25
ma-ke-na si koni
this-CLF:WOODEN-this their sign_of_honour
‘These men beloning to the Tabalu-subclan, this wooden one [i.e. tataba-
board] is their sign of honour’
In sentence (27a), a referent noun, tataba-board, is used without a classifier, while it is
being referred to in (27b) with the classifier ke used for wooden things. The following
examples (28a and b) from Jacaltec (Mayan) demonstrate the use of noun classifiers
as personal pronouns:
(28) Jacaltec (Mayan) (Craig 1986b:264)
a. xil naj xuwan no  lab’a
saw CLF:M.NKIN John CLF:ANIMAL snake
‘John saw the snake’
b. xil naj no 
saw CLF:M.NKIN CLF:ANIMAL
‘He saw it’
Here an interesting generalization of the classifier acquisition studies (Sánchez 1977,
Erbaugh 1996, Senft 1996, Uchida and Imai 1999, Yamamoto 2005) should be
mentioned. Namely, it has been observed that children acquiring a classifier language
seem to be aware of the grammatical function of classifiers at a very young age. This
results in the presence of the basic syntactic pattern of classifier phrases very early on.
Besides, the studies demonstrate that syntactic errors such as a wrong placement of
classifiers are extremely rare in children’s speech. However, due to the heavy
semantic load of classifiers, the semantic criteria for each classifier category are
acquired much later on, resulting in the overwhelming use of general classifiers in the
early stages.

4. Classifier systems of social interaction


As Foley (1997:236) puts it, classifier systems often pay special attention to
categorization of animals and humans “due to the attention they have for us and the
empathetic understanding we hope to share with them.” As he further mentions,
“shape is not the basic semantic parameter here, as this is not what is salient to us
humans when we interact”. It is more important to distinguish, on the one hand,
people from animals, and, on the other hand, to distinguish among humans
themselves, reflecting, for instance, their social or spiritual status.
In this section, I consider several classifier languages which have relatively
large inventories of classifier morphemes used for categorization of human and
animate beings. The classifier inventories of these languages tend to categorize
referents according to a certain social parameter, e.g. status, kinship relation, religious
affinity, age and sex. According to their morpho-syntactic characteristics, the
classifiers represent three different types: numeral, noun and genitive classifiers.

26
The goal is to see whether, and how, these classifier morphemes are anchored to the
social identities of speech act participants, and how, the usage of classifiers varies
depending on the social identities of such. In other words, whether classifiers used for
categorization of human referents have a function of social deixis.

4.1. Genitive classifiers. Social status: Pohnpeian


4.1.1. Pohnpeian
Pohnpeian, a Micronesian language spoken on the Pohnpei island of Micronesia
(formerly written Ponape), has an elaborate system of genitive classifiers. According
to Rehg (1981:179), the exact number of the classifiers present in the language is
difficult to determine. He mentions twenty-one, while Keating (1997:251) speaks of
twenty-nine. What constitutes the major interest for this thesis is the fact that
Pohnpeian system of genitive classifiers has a social status indexing function. As
Keating (1997:247) puts it, “Pohnpeian genitive classifiers not only constitute cultural
categories of rank and power relations, but dynamically re-sort or re-classify these
categories through honorific speech.” Besides this classifier type, the language has a
set of at least twenty-nine numeral classifiers (Rehg 1981:179) denoting primarily
shape but not status. For this reason, the present discussion will be limited to the
system of genitive classifiers.
The natives of the island Pohnpei have historically been organized into more than
twenty matrilineal, totemic, and exogamous clans, and the island is divided into five
independent chiefdoms, with as head a paramount chief and chieftess, and a
secondary chief and chieftess (Garvin and Riesenberg 1952:201, Keating 1997:248).
Today the island is united under a democratic form of government. Despite that, the
chiefdom organization is still vital and remains an important organizing principle of
the Pohnpeian community (Petersen 1982 and Pinsker 1997, cited in Keating
1997:248). As Keating (1997:248, citing Fried 1967) mentions, “there is necessarily a
collaborative construction of leadership in a ranked society, where few instruments
for forcing compliance are available.” The presence of several speech registers
employed in possessive constructions can be explained by the local ideologies about
manaman ‘power’. Mana, a term with cognates throughout the Pacific (Keating
1997:249), or manaman on Pohnpei, denotes the sacred and dangerous power which
flows from the deities through the chiefs to the people. Interestingly, the ‘power’
flows matrilineally to descendants within chiefly clans. It is believed that the
manaman is also transferred from high-status persons to their possessions. As Keating
(1997:248) notes, this “makes possessive constructions a meaningful category in
status-marked relations.”
In Pohnpeian, status indexing can be done by several means, namely: (1) by
the choice of a verb, (2) and/or a noun (there are nouns which index high status, but
almost none which index low status; the exception for the latter is a low-status word
for ‘food’ and a nominalized form of the low-status verb ‘speak’). Additionally, (3)
pronouns can carry a status-indexing role too (the pronouns can be expressed as
exaltive, but not humiliative; according to Keating (1997:249), the common-speech

27
pronouns can therefore be used for the effect of status-lowering in the honorific
speech. Another means, (4) is a special set of greetings. Also, social status can be
encoded by phonological means3 (5). Mainly, however, social indexing is done
through the choice of genitive classifiers (6), which is considered next.
In Pohnpeian, several speech registers can be distinguished, namely:
• common speech vs. honorific speech4, and
• within honorific speech, low status-indexing (humiliative) vs. high status-
indexing (exaltive) registers.
Honorific speech constructs basically two levels of status: one high and one low. That
means that possession in honorific speech is marked either as exaltive or as
humiliative, which is expressed by means of genitive classifiers. Therefore, by means
of classifiers Pohnpeian speakers can either lower their own or another’s status, or
raise another’s. Table 2 gives an overview of the genitive classifiers used in different
speech registers. Examples of the three speech registers with appropriate classifiers
are shown next.

Genitive classifier Speech register Translation


nah pwutak (29) common speech, unmarked for status ‘his/her boy’
sapwellimen Noahs pwutak (30) honorific speech, exaltive (high status) ‘Noah'
s boy’
ah tungoal pwutak (31) honorific speech, humiliative (low status) ‘his/her boy’

Table 2. Genitive classifiers in common, exaltive honorific and humiliative honorific speech (adapted
from Keating 1997:255)

(29) Common speech (Keating 1997:255)


liho pahn kin kadarado
woman.there will habitually send.here
nah kisin pwutako en kohdo en peki.

3 The exaggerated prolongation of vowels in a few common words and expressions can be
parallelled with the honorific speech. According to Rehg (1981:374), in the common greeting
kaselehlie, “the already long vowel eh may be prolonged two to three times its normal length when
addressing a superior or respected equal.” In a similar manner, the vowel e in the word ei ‘yes’ and the
initial vowel ih in ihieng ‘excuse me’, which is used when passing in front of others, are pronounced
longer than normal when one wishes to express respect (Rehg 1981:374).
4 Regarding the honorific speech, it should be pointed that one must differentiate honorific speech
(from Latin honor ‘implying respect’, Thompson (1995:652) and polite speech (from Latin politus
meaning ‘having good manners; cultured; refined, elegant, Thompson (1995:1057). There are certainly
some connections between one and the other; nevertheless, one can be polite in any language (Altman
and Riska 1966, cited in Fischer 1969:418), but not every language has a well-developed system of
honorifics. The distinctive feature of the honorific languages can be their “ability and necessity to
express respect in speech”, in other words “it is impossible to make an utterance of any length in these
languages without openly taking a stand on the amount of respect one is paying to the addressee”
(Fischer 1969:418). For instance, both in Japanese (see Section 2.2.4) and Pohnpeian (the present
section), one proposition can be realized in two or more distinct utterances expressing the same
information but conveying different degrees of respect for the addressee. In a language such as English,
where honorific speech is little developed, most propositions do not have alternative sentences which
will differ from each other exclusively in the degree of respect.

28
CLF:POSS.3SG small boy.there to come to ask
‘That woman always sends her small boy over here to come and ask.’

(30) Exaltive honorific speech (Keating 1997:254-5)


dene Linda oh sapwellimen Noahs pwutak
it.is.said Linda and CLF:POSS.EXAL.of Noah boy
kin patpato rehra (<re ira).
Habitually LocVerb.HML.there location_of_them(D)
‘They say that Linda and Noah’s boy [are] always staying with the two of
them.’

(31) Humiliative honorific speech (Keating 1997:255)


en Saudepe ah tungoalpwutako
of Saudepe CLF:POSS.3SG CLF:POSS.HML boy_there
me ale ira.
the.one take them(D)
‘Saudepe’s boy [is] the one who took the two of them.’

4.1.2. Arguments for the hypothesis


In the present section classifier assignment patterns in honorific speech are considered
and the way the choice of appropriate classifiers varies depending on the social
identities of the speech act participants.
When the speaker refers to the possessions of low-status persons, s/he uses a
humiliative possessive construction. Such constructions are formed on the following
principle: the correct person form of the general classifier a- is followed by the
humiliative classifier tungoal. Thus, the specific classifiers which are usually used
with nouns in common speech are replaced by the general plus the humiliative
classifier.

Common speech Humiliative honorific speech5 Translation


kene mwahng ah tungoal mwahng ‘her taro’
nah pwihk ah tungoal pwihk ‘her pig’
were sidohsa ah tungoal sidohsa ‘her car’

Table 3. Genitive classifiers in common and humiliative honorific speech (adapted from Keating
1997:255)

Interestingly, by means of the a- tungoal formation and the concurrent use of two

5 Unlike in common speech (and exaltive honorific speech), in humiliative honorific speech the
categorization between different notions, for example such as a vehicle and food, is not observed. In
common speech not only these notions are distinguished from each other, but for instance, between
food itself. For instance, food gets different classifiers depending on whether it is about to be eaten or
still on the tree (Keating 1997:255).

29
classifiers, the speaker can express two different levels of status in one construction
simultaneously:
(32) High and low status in the same construction (Keating 1997:257)
omw-i tungoal moahng
your.SG-[EXAL] CLF:POSS.HML head
‘your.SG head’
One level of status is expressed by the first classifier, the other level of status is
communicated by the second one, since the exaltive suffix -i signals high status while
the humiliative form tungoal signals low status (Keating 1997:256-7). Using this
construction the speaker can ensure that an addressee and a bystander, or an
addressee and a referent, can be indexed as having two different statuses.
In the situation given below, one low-status man is offering a cigarette to
another in the presence of a chief. Both the speaker and the addressee are of lower
status than the chief, and this is reflected in the choice of classifiers.
(33) High and low status in the same construction (Keating 1997:257)
Nanihdpei omw-i tungoal sikah
Nanihdpei your.SG-[EXAL] CLF:POSS.HML cigaret
‘Nanihdpei. Your cigarette (offering him a cigarette).’
Example (33) shows that the speaker expresses deference to the addressee (Nanihdpei,
in this case), yet he acknowledges the status of another higher-status by adding
tungoal to the utterance.
When the speaker refers to anything owned by a high-status person, s/he uses exaltive
possessive constructions which includes a special set of genitive exaltive classifiers.
As Keating (1997:258) mentions, some exaltive possessive constructions might be
formed without the use of classifiers but by suffixing the possessive pronouns to the
honorifically marked noun denoting the thing possessed. She adds that some of
Rehg’s (1981) consultants reported that nouns referring to chiefs do not normally take
possessive classifiers at all, since the honorific nouns themselves can communicate
high status of the possessor. The interactional data of Keating shows, however, a
different situation. Namely, exaltive possession is often expressed with both a
classifier and an honorific noun6. Example (34) illustrates the case with the beginning
of a man' s prayer at a feast.
(34) Exaltive possession construction (Keating 1997:259)
maing samaht Koht (…)
honoured_one father_our(EXCL) God
iangaki sapwellimomwi kahlap Isipahu,
join_with CLF:POSS.2SG.EXAL body[EXAL] paramount_chief

6 In the interactional data recorded by Keating (1997), the exaltive possessive classifiers tehnpese
and tehnwere expressed in the forms of tehnpas ‘dwelling[EXAL]’ and tehnwahr ‘vehicle[EXAL]’,
appear without possessive marking, “since they specifically indicate the chief’s possessions already”
(Keating 1997:259).

30
Iso Nahnken, Likend
secondary_chief paramount_chieftess
‘Our honoured father God, join with your body the paramount chief, the
secondary chief, the paramount chieftess...’

As can be seen, the second line of example (34) contains a double honorific marking,
due to the use of the exaltive possessive classifier sapwelliwe together with the
honorifically marked noun for body kahlap. Keating (1997:259) draws attention to the
fact that the construction maing samaht Koht in the first line of example (34) is
unmarked for status. She adds that this is a common speech part-whole construction
which is highly conventionalized in prayer openings.
The following example offers another case of redundant exaltive marking. It is
an abstract of a speech of a medium-ranking woman at a feast in which she refers to
two different statuses of spouses by choice of an exaltive vs. humiliative classifier and
by choice of an exaltive vs. common noun.
(35) Exaltive possession construction (Keating 1997:259)
Patpato tikitikieng ohng eh
LOC.Verb[HML].there small_to for uh
Sapwellimatail werek kan de
CLF:POSS.1.INCL.EXAL spouse[EXAL] those or
atail tungoal pwoud kan
CLF:POSS.1.INCL CLF:POSS.HML spouse those_by_you
‘Talking a little bit (i.e. nicely) to uh our high-status spouses or our low-status
spouses.’
Similarly, in the second line of example (35), both the noun werek ‘spouse’ and the
classifier sapwellime are indexed for status. At the same time, the status is
communicated only by the classifier in the humiliative construction in line three.
Summarizing, in Pohnpeian the usage of genitive classifiers is determined on basis of
the relative social position of speech act participants. Depending on the relative
position of the speaker to an addressee, a referent or a bystander, the choice of a
classifier will vary. Interestingly, some constructions even enabled the speaker to
express two different levels of status simultaneously. As shown in examples (32) and
(33), the speaker, an addressee, a bystander or a referent can be indexed as having two
different levels of status.
In this respect the inventory of genitive classifiers in Pohnpeian makes part of
a shifting set of relationships, depending on who is speaking to whom and who is
listening in on the conversation. Therefore, I argue that the considered genitive
classifiers represent socially deictic items, since they anchor "to the social identities of
participants […] in the speech event, or to relations between them…” (Levinson
1979:206).

31
4.2. Noun classifiers. Age and kinship: Jacaltec, Mam and Akateko
The present section deals with an intersecting area of classifiers and such parameters
as kinship, sex and age. It is of interest to consider how these social parameters can be
brought in focus by classifiers and how the choice of an appropriate classifier will
vary depending on age, sex and kinship relation of the participants of a speech event.
Inventories of classifiers characterizing noun referents according to kinship,
sex and age have been attested in the Mayan language family. In Section 4.2.1, the
noun classifiers encountered in Jacaltec are considered. Section 4.2.2 focuses on the
classifier system of another Mayan language - Mam. In Section 4.2.3, the noun
classifiers in Akateko, yet another Mayan language, are discussed very briefly.

4.2.1. Jacaltec
Jacaltec7 belongs to the Kanjobalan branch of the Mayan language family (Craig
1986b:263). It is spoken in a portion of the north-western slope of the Cuchumatanes
mountains in the department of Huehuetenango, Guatemala, and in several nearby
settlements in the Mexican state of Chiapas (Day1973a:3). There are several
dependent settlements in each Jacaltec-speaking town. The people of such settlements
usually speak the dialect of their town and consider themselves members of the town
society. According to Day (1973a:4-5), virtually all Jacaltec speaking dwellers of the
town of Jacaltenango, whom he interviewed, “have close family ties in the town and
return to it now and then for Sunday markets, major town fiestas of close town
relatives.” This is one of the features of the culture which echoes in the system of
classifiers (categorization of nouns according to kinship, sex and age). Among the
neighbouring languages of Jacaltec, Mam should be mentioned, which is spoken in
the area to the south and the Kanjobal language to the east and north-east. To the
north and west the area is bordered by Spanish speaking Ladino towns (Day 1973a:5).
Jacaltec has several classifier systems. One is the system of numeral
classifiers, which is referred to in Craig (1986b:265) as ‘number classifiers’. This very
limited inventory categorizes all nouns according to three parameters: humanness,
animacy, and non-animacy. Another system which Craig (1986b:265) mentions
referring to Day is the ‘numeral classifier’ system, which, in fact, is a system of
‘measure terms’ as defined in Section 3.1, according to Grinevald’s (2000) nominal
classification continuum. The third classifier system (or, actually, the second, if we do
not take measure terms into consideration) is the system of noun classifiers. For this
thesis the latter constitutes major interest and is considered in more detail in the next
subsection.
The Jacaltec noun classifier system consists of twenty-four classes/categories
which are all semantically transparent. Craig (1986b:266) gives two reasons for this.
First, this noun classifier system can be regarded as frozen: new loanwords are not
classified, hence the system does not expand (Craig 2000, cited in Aikhenvald
2000:313). Second, there is no residual or general category. The given noun classifier

7 Jacaltec is also known as Popti’, the name used by the speakers themselves in reference to their
language.

32
system can be divided into two subsystems: one categorizes the humans, the other the
non-human world. Craig (1986b:266), following Denny (1976), characterizes the
domain of the former as that of social interaction and the domain of the latter as that
of physical and functional interaction. The noun classifiers of the former subsystem
are presented in Table 4.

Noun classifier Semantics


1. cumam male deity
2 cumimyi  female deity
3. ya  respected human
4. naj male non-kin
5. ix female non-kin
6. naj ni  an young male non-kin
7. ix ni  an young female non-kin
8. ho  male kin
9. xo  female kin
10. ho  ni  an young male kin
11. xo  ni  an young female kin
12. unin infant

Table 4. Jacaltec noun classifiers. System of social interaction (adapted from Craig 1986b:266-7)

Here it should be noted that the sources offering the list of classifiers (Day 1973b) and
Craig (1986b) slightly differ not only in the spelling/transcription, but also in
semantics of some classifiers. Craig (1986b:268) mentions that the Jacaltec classifier
system of social interaction was originally described in Day (1967). In his definition
of classes, Day isolated such semantic parameters as divinity, kinship, respect, sex
and age. Craig (1986b) presents the list of classifiers which categorizes humans also
according to the parameters of divinity, kinship, respect, sex and age. However, in
Craig’s categorization, (a) the parameter of age has no age gradations (e.g. non-infant
vs. youth vs. non-child) as in Day (1973b), and (b) the classifiers used for adults are
unmarked for respect. The list of noun classifiers given in Table 4 are borrowed from
Craig (1986b), the explanatory notes which follow are found in Day (1973b:87).
The noun classifiers cumam (1)8 and cumi  (2) (in Day 1973b:86, komam and
kom  , respectively) originate from cu man and cu m  ‘our father’ and ‘our mother’
(Day 1973b:86, Craig 1986b:271). The classifier for male deity is used for the
Spanish loans dyos ‘God’, xesukrísto ‘Jesus’, which are regarded as high male deities.
The only high female deity is šahanlax mi  ‘the Virgin Mary’. There are a whole
number of low deities, which can be used either with these two classifiers or with naj
(4) and ix (5). Male low deities include male tiyoš ‘saints’, k’uh ‘lightning bolts’, and
diseases such as ‘measles’, ‘smallpox’, and ‘whooping cough’. ‘The moon’, ‘the corn’
and female ‘saints’ are female low deities. Day (1973b:87) notes that “…“Deity” is
almost certainly a misleading term, but must serve without explanation until better

8 For the sake of ease, the classifiers in the explanatory notes are provided with the number
according to which they appear in Table 4.

33
ethnographic data are available.” One can use komam for pale Catholic ‘priest’, but
Day’s informant regards it as an exaggerated and little used way of showing respect.
Normally, priests are classified as ya  (3).
The noun classifier ya  (3) is used for any respected person that is not a deity.
While age difference partly determines respect of one person for another, some other
factors besides age can also be important. As Day (1973b:87) mentions, “for priests,
age may have no bearing at all.”
Naj (4) and ix (5) are used for persons, which are not included in all other
categories. These classifiers are also applied to dead people and to Ladinos (which for
Jacaltec speakers includes North Americans).
Ho  (8) and xo  (9) are applied to non-infant kin. According to the informant
of Day, the age of non-infants is regarded as approximately 1,5 to 10 years. The data
given by Craig (1986b:267) differs slightly with respect to the age gradations. She
states that ho  (8) and xo  (9) are used with adult kin. Here it is of interest to consider
the general tendency to associate certain age groups with non-infant children, youth
and adulthood. Different perception of the age categories by the informants of Day
and Craig could be the reason for the slight difference in data. It should be also
mentioned that, according to Day (1973b:87), the kinship parameter in Jacaltec has
“maximum horizontal extension to first cousins and siblings-in-law, and vertical
extension to the second generation”. Craig (1986b:271), however, has a different
standpoint. She notes that “the use of the kin classifier is reserved to close members of
the family and does not extend to members of an extended family.” The classifiers
ho  (8) and xo  (9), according to Day (1973b:87), can also be used for non-infant
godchildren.
Unin (12) is used for all infants. The age approximation of an infant is from
birth to one-and-a-half years according the Day’s informant.
Compound noun classifiers with ni  an ‘little’, (6), (7), (10) and (11), are used
for non-infant children. Craig (1986b:266-7) labels ni  an with the general term
‘young’. The use of it is optional for kin children, but obligatory for non-kin ones.
The following example demonstrates the use of the classifiers naj ‘male, non-
kin’ and ya  ‘respected person’:
(36) Jacaltec (Mayan) (Craig 1986b:264)
xul naj Pel b’oj ya  malin
came CLF:M.NKIN Peter with CLF:RESPECTED.PERSON Mary
‘Peter came with Mary’
The structure of the given classifier inventory is shown in Figure 3. As can be seen
from the figure, deities and supernatural entities are treated apart from the human
world. They are subcategorized into female and male deities. Further categorization
follows according to the kinship parameter. This parameter is not relevant for infants
and persons of respect. Craig (1986b:268) mentions that “although respect may
naturally coincide with old age, the semantic feature of respect is independent of the
one of age”, and that “Jacaltecs say of persons classified as ya  that they have
acquired respect through various means, only one of which is the life experience and
wisdom that may accompany old age.” The other eight classifiers not only categorize

34
according to kinship, but also according to sex and age.

Figure 3. Jacaltec classifiers of social interaction (adopted from Craig 1986b:268)

4.2.2. Mam
Mam belongs to the Mamean branch of the Mayan language family. It is spoken in the
highlands of western Guatemala. Table 5 shows the list of Mam noun classifiers that
is found in England (1983). According to England (1983:158), the noun classifiers are
used “when third person noun phrases are deleted from a sentence.”

Classifier Semantics
1. jal non-human
2. ma man
3. xu  j woman
4. nu  xh baby
5. xhlaaq’ child
6. q’a young man
7. txin young woman
8. swe  j old man
9. xhyaa  old woman
10. xnuq old man, respectfully
11. xuj old woman, respectfully
12. b’ixh person of the same status, fondly
Table 5. Noun classifiers in Mam (adapted from England 1983:158)

It can be seen from the table that there is only one classifier used for non-humans jal
(1). The other eleven classifiers which categorize the human world, can be
characterized as sensitive to such semantic parameters as age (baby/child/young
person/old person), sex (M/F) and respect (marked vs. unmarked for respect). Babies

35
and children are not further subdivided according to sex, while the young people and
the old are. There are two classifiers which include respect in their semantics: xnuq
(10) and xuj (11) for old man and old woman, respectively. There is also one classifier
used for persons of the same status. The following example illustrates the classifier
usage in Mam. In sentence (37a) the noun referent ‘my father’ is introduced without a
classifier, in (37b) it is referred to only with the classifier swe  j ‘old person’, thus
functioning as a personal pronoun:
(37) Mam (Mayan) (England 1983:159)
a. aa-Ø-tzan n-maan-a aj q’i-1 t-wittz-xax
DEM-3SG.ABS-well 1SG-father-1SG REL give-AG 3SG-head-ENCL
‘It is my father who plans’
b. aa-Ø-swe  j aj
q’i-1 t-wittz-xax
DEM-3SG.ABS- CLF:M.OLD give-AG 3SG-head-ENCL
REL
‘It is he (older person) who plans’
To make the overview more illustrative, in Figure 4, the classifiers are rearranged
along the dimensions of age, respect and sex. Thus, the horizontal axis accounts for
the parameter of sex, while the vertical axis shows the division into classes according
to age, status, respect, and humanness.

Non-humans ----> jal

{
b’ixh <-- (person of the same status)
M <------ Sex ------> F
Respect, status
xnuq xuj <-- (old person, respectfully)
M <------ Sex ------> F
swe  j xhyaa  <---- (old person)
ma xu  j <---- (unmarked for age)
q’a txin <---- (young person)
xhlaaq’ <---- (child)
nu  xh <---- (baby)

Figure 4. Mam noun classifiers along the dimensions of age, respect and sex (structure of the table is
inspired by Craig 1986b)

4.2.3. Akateko
Akateko belongs to the Kanjobalan branch of the Mayan language family. It is spoken
in Guatemala and encounters about 25,000 speakers. The noun classifiers in Akateko
have a slightly different semantic content.
According to Mateo Toledo (1999:150), the noun classifiers in Akateko used with
human referents are as follows. For female referents the form ix is used, for male
referents the naq form. However, in various Akateko speaking areas, these two
classifiers have a slightly different semantic content. In San Rafael and Acatán, in
some cases, they are used to refer to a man or a woman that are unfamiliar to the

36
speaker. In Ixcoy, they are used exclusively to refer to a man or woman that are
unfamiliar. In other places, the usage of the terms is independent of the fact whether a
female or male referent is familiar to the speaker or not. Example (38) demonstrates
the use of the noun classifier naq in the Akateko speaking area Ixcoy.
(38) Akateko (Mayan) (Mateo Toledo 1999:150)
naq winaq
CLF:M.UNKNOWN man
‘a man’
Concerning the forms which are used in reference to persons of great age or to
persons of a high social status (in terms of profession, marital status, etc.) the
following information is available. In Soloma and Ixcoy, the term xala is used in
reference to a respected woman, realized as xal in Santa Eulalia and Barillas. In
reference to a respected man, in all for areas, the term cham is used. In San Rafael and
Acatán these classifiers are not used, but the terms ix and naj, mentioned in the
beginning, are employed in reference to a female or male person, with no distinction
made whether they are respected or not. The usage of these classifiers is extended to
personification / categorization of super-natural elements, illnesses, inexplicable
events, etc. Example (39) illustrates the use of the respect classifier xal encountered in
the areas of Santa Eulalia and Barillas (Mateo Toledo 1999:150).
(39) Akateko (Mayan) (Mateo Toledo 1999:150)
xal malin
CLF:F.RESP María
‘señora María’
The Akateko noun classifiers used with humans are summarized in Table 6.

ACA/RAF IXC/SOL EUL/BAR Semantics of classifiers


1. naj naq naq man
2. ix ix ix woman
3. ix ix ix child
4. - cham cham man, respectfully
5. - xala xal woman, respectfully
6. ya ya - man, fondly
7. k’o q’o - woman, fondly
9
Table 6. Noun classifiers in the Akateko speaking areas .

Mateo Toledo (1999:150) also mentions special classifier forms which have a
connotation of fondness. Such classifiers are used in reference to people whom one is
fond of. Thus, in the area of Ixcoy, the speaker uses the form ya and the form q’o
referring to, respectively, a male or female whom s/he likes. In Acatán and San Rafael
these classifiers are also used (the form for female referents realized as k’o), but the

9 The abbreviations for the areas are as follows: ACA – Acatán, RAF - San Rafael, IXC – Ixcoy,
SOL – Soloma, EUL - Santa Eulalia, BAR – Barillas (adopted from Mateo Toledo 1999:150).

37
usage is not consistent. In other areas these classifiers are not employed.

4.2.4. Arguments for the hypothesis


The present section focuses on the usage patterns of the noun classifiers in Jacaltec,
Mam and Akateko and the ways in which the choice of a classifier varies depending
on the social identity of the speaker, addressee or a referent in a speech event.
In Jacaltec, humans are categorized according to age, sex, kinship relation, status and
religious affinity. In order to see whether and how the choice of the appropriate
classifier depends on the above mentioned parameters, a standard classifier
assignment scheme can be considered. Thus, for instance, a middle-aged man with a
certain status in his own community can be referred to with different classifiers. By
the man' s contemporaries and elder persons he could be referred to as naj; by his
close relatives as ho  ; by someone younger or by contemporaries who are not close
friends or neighbours as ya  (Craig 1986b:270).
Figure 5 shows the internal organization of the Jacaltec classifier inventory
and schematizes the standard classifier assignment process. The classifiers are ordered
along two dimensions on the figure: the horizontal axis accounts for the parameters of
kinship and sex, and the vertical axis accounts for the division into classes according
to such parameters as divinity, respect and age. The choice of an appropriate classifier
will vary according to the position of the speaker relative to the referent, their kinship
relation, and according to certain properties of the referent, for instance, religious
affinity, sex and age.

M <-- sex --> F


Divinity ---------------------> cumam - cumi 

Respect ----------------------> ya 

Non-kin <-- Kinship --> Kin


M <-- sex --> F M <-- sex --> F
naj ix ho  xo  <---- (adult)
naj ni  an ix ni  an ho  ni  an xo  ni  an <---- (child)

unin <----------------------- (infant)


Age

Figure 5. Internal organization of the classifiers of social interaction (adapted from Craig 1986b:269)

Here it must be noted that a speaker is able to convey pragmatic meaning varying the
choice of a classifier. In what follows such cases are considered.
As mentioned before, the classifiers naj and ix are used as a kind of residual
category that is also applied to dead people and Ladinos. Interestingly enough, Day
recorded the use of another classifier in address to Ladinos, namely the classifier ya 
(see Table 4), which is normally used for respected humans. Thus, as Day (1973b:87)

38
mentions, ya  can be accorded to non-Indians on some occasions, for example, to a
“Jacaltec-speaking Ladino employer in his presence”. This way a speaker is able to
communicate respect or show solidarity by treating a Ladino as insider of the Indian
community. Such re-classifications are possible along the two dimensions shown in
Figure 5. For instance, along the respect and age dimension, “one may express extra
positive or extra negative feelings toward the person referred to by bestowing or
withholding expression of respect” (Craig 1986b:270). It is insulting for a person of
great age or fame to be referred to with the naj and ix classifiers. The opposite is also
true, as already shown above: for someone who is expected to be classified as naj and
ix it is a form of compliment to be referred to as ya  . Day (1973b:88) notes an
interesting fact, namely, in order to express affection and love to a child or youth, an
adult may refer to him/her with the classifier unin, usually used for infants, thus
replacing the ‘standard’ in this case classifiers naj ni  an or ho  ni  an or their feminine
counterparts. One can also express extra affection by ‘playing’ with classifiers along
the kinship dimension. Thus, referring to a favourite non-kin using classifiers for kins,
one can communicate fond and kind feelings. Conversely, to express anger in
reference to a kin relative one can use classifiers for non-kins (Craig 1986b:270). The
list of classifiers given in Day (1973b:86) shows, that the semantics of naj and ix
classifiers include the ‘indulgently loved’ senses. Surprisingly enough, naj and ix,
being residual categories, which categorize not only male and female non-kins, but
also dead people and Ladinos, can nevertheless be used to communicate a great
amount of affection by an adult for a child. Day (1973b:88) notes that the classifiers
naj and ix “may be used for a kin or non-kin infant or non-infant child, or a kin youth”
instead of unin, naj ni  an, ho  ni  an, and ho  , or their feminine equivalents. No data
is available whether adults can be reclassified and referred to with ‘indulgently loved’
forms in order to express affection of an adult for his/her contemporary.
As can be seen, the system of the noun classifiers in Jacaltec is quite dynamic.
A human referent is seldom associated with one particular classifier. Depending on
social status, religious status, sex of a referent, his/her age relative to the speaker and
kinship relation with the speaker, the choice of a classifier will vary. The choice of a
classifier, however, does not always reflect the real social relationship. The speaker
can vary with assigning a classifier in order to create pragmatic focus and
communicate his/her own estimate of the social relationship between him/her and a
referent. Therefore, taking the process of classifier assignment in Jacaltec into
account, it can be argued that the noun classifiers in Jacaltec have the function of
social deixis, for they anchor "to the social identities of participants […] in the speech
event, or to relations between them…” (Levinson 1979:206).

Mam, one of the other Mayan languages considered, has an inventory of twelve noun
classifiers, eleven of which are used for human referents. The referents are
categorized mainly according to age and sex, though three classifiers from the given
inventory, xnuq, xuj and b’ixh, include the parameter of respect to their semantics.
The latter classifiers are able to reflect social position of a referent which does not
always coincide with great age. Younger speakers, for instance, are likely to use the

39
forms xnuq and xuj in reference to older people. Nevertheless, people of great status,
like teachers, are referred to by these terms as well (England, p.c.10). The classifier
b’ixh is employed for reference to people who share the same status in the
community. There is hardly any information available on the dynamic qualities of the
Mam classifier system. However, taking into account the categories these classifiers
form, the following assumptions can be made:
• A speaker can vary in assigning respect. Thus one might use the respectful
xnuq and xuj classifiers for someone for who the standard swe  j and xhyaa 
could be sufficient, communicating in this way his/her deference.
Alternatively, it might be a form of insult for an elder person of respect to be
referred to with the classifiers swe  j and xhyaa  , which have no ‘respect’
connotation.
• A speaker can vary with the use of the b’ixh classifier, thus expressing either
solidarity or social distance.
If these statements were correct, the choice of appropriate classifier forms would be
determined by the social position of the speaker relative to the position of a referent.
For instance, reference to a male person would be possible by various classifiers
depending on who is speaking to whom and/or what the pragmatic focus is: if the
speaker and a referent share the same status in the community, the term b’ixh can be
used; if the speaker is of a lower status or s/he wants to communicate respect, then
xnuq is more appropriate; with the term swe  j no reference is made to the social
position of the referent but only to his/her great age. Thus, these assumptions, if
supported by empirical data, would provide arguments in support of the hypothesis
that the classifiers in question have a socially deictic function.
However, those classifiers which make reference exclusively to sex of a
referent, without including the parameters of respect and status in their semantic
content, do not seem to qualify for socially deictic items. The choice and assignment
of such classifiers would not require a relative perspective and would not vary from
speaker to speaker. In other words, the assignment of the classifiers categorizing
referents exclusively according to sex is not determined by the social identity of the
speaker and his/her relation to the referent, as it is determined by sex of the referent
only. A quite debatable question is whether the use of classifiers categorizing human
referents according to age can depend and reflect information on the social identity of
the speaker. Example (37a), repeated in (40a), and example (40b) from Mam
demonstrate the use of noun classifiers making reference to sex and age:
(40) Mam (Mayan) (England 1983:159)
a. aa-Ø-swe  j aj q’i-1 t-wittz-xax
DEM-3SG.ABS- CLF:M.OLD REL give-AG 3SG-head-ENCL
‘It is he (older person) who plans’

10 Many thanks to Dr. M.Crevels for communicating personally with Prof. Dr. N.England on this
question.

40
b. n-Ø-  aq’naa-n malin t-uuk’al-q’a
PROGR-3SG.ABS-work-AP María 3SG-RN/COM-CLF:M.YOUNG
‘María works with him’
In example (40a), the classifier swe  j ‘old person, male’ refers to the noun ‘father’
introduced earlier in the discourse. In this case the choice of the given classifier is
likely to have been determined by the relative standpoint. In example (40b) it is
unclear though, whether the choice of the classifier q’a ‘young person, male’ refers to
the ‘absolute’ (if possible at all) or relative age of the referent.

As for Akateko, the semantic content of noun classifiers for humans in this language
is slightly different in various Akateko speaking areas. While in some areas, the noun
classifier forms ix and naq are used simply for female and male referents, in other
areas these forms are used for female and male referents which are unfamiliar to the
speaker. Akateko also has noun classifiers with a semantic content of respect. These
are employed in reference to persons of great age or to those of a high social status (in
terms of profession, marital status, etc.). The usage of these classifiers differs from
area to area. Example (39), repeated in (41), illustrates the use of the ' respect'
classifier xal, encountered in the areas of Santa Eulalia and Barillas (Mateo Toledo
1999:150).
(41) Akateko (Mayan) (Mateo Toledo 1999:150)
xal malin
CLF:F.RESP María
‘señora María’
The case of Akateko also shows that the choice of the classifiers that categorize
humans according to respect/status is determined by the position of the speaker
relative to the referent and/or by the social relation between them. Thus, the social
identities of the speaker and the referent get encoded in the classifier forms. This
characteristic allows us to regard these classifiers as socially deictic items. The choice
of classifiers which categorizes a human only by sex or age does not depend on any
other factor than sex or age of the referent, thus reflecting no information on other
speech act participants, for instance, on the speaker. Such classifier forms do not
qualify for socially deictic items according to the followed definition of social deixis.
However, in some Akateko speaking areas, the latter classifiers are used exclusively
with referents which are unfamiliar to the speaker; thus, encoding information on the
social relationship between the speaker and the referent in a certain way.

4.3. Numeral classifiers. Family membership: Yi branch languages


This section considers languages of the Yi branch of Burmic within the Tibeto-
Burman language family. The Yi branch languages are reported to have specific
collective numeral classifiers for family groups. This unique feature makes it
interesting to consider these languages in search for additional arguments for the
hypothesis. Additionally, these classifiers are morpho-syntactically distinct from other

41
attested classifiers, since most of them are disyllabic comprising two morphemes
(Bradley 2001:1).
Yi branch languages are also known as Lolo or Loloish, though the Yi people
regard this as a pejorative name. Concerning the geographic distribution, the Yi
branch languages are spoken in south-western China and adjacent South-east Asia.
More precise locations and speaker numbers are given in the next subsections as soon
as particular languages of the branch are introduced.
To illustrate the phenomenon of numeral classifiers in Yi languages, the following
example can be given. Both noun phrases (42a and b) have a general meaning ‘three
people’, but (42a) contains a monosyllabic classifier for human referents while (42b)
exemplifies a family group classifier with a specific meaning ‘father and sons’.
(42) Lipo (Central Yi, Tibeto-Burman) (Bradley 2001:2)
a. tsha33 so33 jo33
person three NUM.CLF:HUM
‘three people’
b. so33 po55lo21
three NUM.FGCL:FATHER
‘a father and two sons’
The number, ‘three’ in this case, refers to the total number of people and not
specifically to the number of children. Bradley (2001:3) mentions the following
psychological phenomenon: when speakers of a language like English “are asked how
many brothers or sisters they have, they do not include themselves in the answer; but
speakers of these languages do; as do the numbers accompanying family group
classifiers.”
What follows is an illustration of a wide range of family group classifiers
existing in the languages of the Yi branch. Thus, not every language of the Yi branch
is considered and not the entire inventories of family group classifiers of the
mentioned languages are given (see Bradley 2001). The aims are (a) to demonstrate
the possible semantic range of family group classifiers and (b) to see whether the
properties of the family group classifiers in these languages will provide any
argument for the hypothesis.

4.3.1. Akha and Hani (Southern Yi languages)


Akha, a language of the Southern sub-branch, counts about 600,000 speakers. It is
mainly found in the extreme southwest of Yunnan in China, in north-eastern Burma,
in Thailand and Laos (Bradley 2001:3).
Table 7 shows kin terms for mother, father and grandparents in Akha and
corresponding family group classifiers.

42
Kin terms Family group classifiers
21 33 21 33 21
a da ‘father’ + za ‘child’ -> da za ‘father + child’
21 33 21 33 21
a ma ‘mother’ + za ‘child’ -> ma za ‘mother + child’
21 55 21 55 21
a b ‘grandfather’ + za ‘child’ -> b  za ‘grandfather + grandchildren’
21 21 21 21 21
a pi ‘grandmother’ + za ‘child’ -> pi za ‘grandmother + grandchildren’

Table 7. Kin terms and family group classifiers in Akha (adapted from Bradley 2001:3)

As can be seen in Table 7, the family group classifiers in Akha have a transparent
structure. The first syllable is derived from the relevant bound kin term and the
second syllable is a noun with a meaning ‘son’ or ‘child’. Bradley (2001:4) gives the
following semantic generalization:
a person of the highest generation and appropriate gender and at least one of the lowest
generation and either gender must be included. Others of any included generation of either
gender may be included […]. Those included must all be from the same lineal family, and
are usually from one nuclear family plus patrilineal or matrilineal grandparents; affines can
be included.
The statement is exemplified by (43a and b). Note the impressive list of possible
meanings of the phrase in (43b).
(43) Akha (Southern Yi, Tibeto-Burman) (Bradley 2001:4)
21 33 21
a. sm ma za
three NUM.FGCL:MOTHER
‘mother and two children’ (or ‘mother, father, and one child)
21 33 21
b. ø pi za
four NUM.FGCL:GRANDMOTHER
‘grandmother and three grandchildren’ (or ‘grandmother, grandfather and two
grandchildren’, or ‘grandmother, daughter and two grandchildren’, or
‘grandmother, grandfather, daughter and one grandchild, or ‘grandmother,
grandfather, son and one grandchild’, or ‘grandmother, son and two
grandchildren, or ‘grandmother, son, son’s wife and one grandchild, or
‘grandmother, son’s wife and two grandchildren, or ‘grandmother, daughter’s
husband and two grandchildren’ or ‘grandmother, daughter, daughter’s
husband and one grandchild, or ‘grandmother, two grandchildren and one
grandchild’s spouse, etc.)
In addition to the already mentioned family group classifiers, Akha has a family group
classifier for siblings (m 55nm55) (Bradley 2001:4):
21 55 55
(44) sm m nm
three NUM.FGCL:SIBLINGS
‘three siblings’
Hani is a language found in south-central Yunnan and adjacent areas of Vietnam with,
in total, about 1.350,000 speakers. Hani employs morpho-syntactically similar family
group classifiers as exemplified for Akha in Table 7. Besides the already mentioned

43
classifiers, Hani also has the following ones (Bradley 2001:5):
55 21
d i za
 ‘spouses’
21 21
da za ‘group of young people’
33 21
tsa za ‘ethnic group’

4.3.2. Nasu and Nosu (Northern Yi languages)


According to Bradley (2001:5), Northern Yi languages are characterised by a
relatively elaborated kinship system and a larger inventory of disyllabic family group
classifiers for sets of kin than in the Southern and Central languages. It must also be
mentioned that in the Northern Yi languages, the family group classifiers have
different compounding and slightly different semantics in comparison to the Southern
languages.
This section refers to two languages of the Northern group, Nasu and Nosu.
Nasu is spoken “by several hundred thousand members of the Yi nationality in north
central Yunnan in China, especially in Luquan and Wuding counties and also in
surrounding areas” (Bradley 2001:5). The language is also known in Chinese as Hei
Yi (‘Black Yi’). Nosu has the largest number of speakers of the languages of the
Northern branch, it is spoken by nearly three million people. It is found in south-
western Sichuan and north-western Yunnan.
Table 8 presents the family group classifiers with one- and two-generation
difference attested in Nasu.

Family group classifiers Semantics


h 33 33
po zo ‘father + sons’ (only males are included)
21 33
mo zo ‘mother + one or more children’
p i 

h 55 33
‘grandparent(s) + grandchildren’

Table 8. Family group classifiers in Nasu (adapted from Bradley 2001:6)

The first two do not contain the kin terms for father and mother, but do contain the
morphemes with semantics ‘male’ and ‘female’ (pho33 and mo21 respectively).
Interestingly, in Nasu, the family group classifier for ‘father and son’ cannot include
daughters, while the one in the other languages can. The family group classifier for
‘mother and children’ is therefore semantically richer: “it can even be used for a
father plus his children including at least one daughter but not necessarily including a
mother” (Bradley 2001:6).
Additionally, Nasu has the following four disyllabic and monosyllabic family
group classifiers which are interesting from the perspective of their semantic content.
The disyllabic ones are:
(i) (me55) t he21, with meanings:

- ‘potential wives for males’
- ‘potential brothers’ wives for females’
21
(ii) zo55, with meanings:
- ‘brothers of potential wives for males’

44
- ‘brothers of potential brothers’ wives for females’.
The monosyllabic forms are:
(iii) f 33, including

- “all non-marriageable members of the same patriclan in one’s own
generation, male or female, with the same father or not” (Bradley 2001:7),
(iv) pho55, with meanings:
- ‘father and sons’
- ‘father’s brothers’ sons’,
The latter two meanings include males only. The two disyllabic and two monosyllabic
family group classifiers are summarized in Table 9.

Family group classifier Semantics


55 h 21
(me ) t e matrilateral female cross-cousins or brother’s wives
21 55
zo matrilateral male cross-cousins or wives’ brothers
h 55
po male parallel cousins or father and sons
33
f siblings and patrilateral parallel cousins (either gender)

Table 9. Additional family group classifiers in Nasu (adapted from Bradley 2001:7)

Nosu, another Northern Yi language, has one family group classifier which should be
mentioned additionally. This is 5 i33, the disyllabic family group classifier for

brothers:
(45) Nosu (Northern Yi, Tibeto-Burman) (Ayu 1989:13, cited in Bradley 2001:8)
so33 i
5 33

three NUM.FGCL:BROTHERS
‘three brothers’

4.3.3. Lisu (Central Yi language)


Lisu, a Central Yi language, counts over a million speakers in north-western Yunnan
and in Burma, Thailand, India and Laos. Besides a full four-member disyllabic family
group classifier paradigm, Lisu has some monosyllabic family group classifiers which
are of interest to consider. Bradley (2001:12) mentions l55, the family group 
classifier which can be used a) in reference to groups of same-generation siblings and
cousins, or b) in reference to spouses. Which of the two meanings is more appropriate
is partly determined by the number with which the family group classifier occurs.
Thus, with the number ‘two’, the more likely meaning is ‘spouses’. This also applies
to even numbers from ‘four’ up, since it is possible that one refers to more than one
married couple. The reading ‘sibling’ or ‘cousin’ is not unlikely either with even
numbers from ‘four’ up. With odd numbers from ‘three’ up, the meaning of ‘sibling’
or ‘cousin’ is very likely but not the one of ‘spouses’, “except for polygamous
families” as Bradley (2001:12) notes. The use of this family group classifier is
exemplified below.

45
(46) Lisu (Central Yi, Tibeto-Burman) (Bradley 2001:12)
55
a. wa33nu21 i21
  l
we(EXCL) two NUM.FGCL:SIBLING.SPOUSE
‘The two of us’ (spouses, siblings or cousins)
55
b. wa33nu21 sa44
 l
we(EXCL) two NUM.FGCL:SIBLING.SPOUSE
‘The three of us’ (siblings or cousins)
55
c. wa33nu21 li44
 l
we(EXCL) four NUM.FGCL:SIBLING.SPOUSE
‘The four of us’ (siblings, cousins, or two pairs of spouses)
Disyllabic family group classifiers in Lisu can also have a whole range of extended
meanings. Bradley (2001:13) mentions that many Lisu speakers lack the generalized
form mi55li21 (‘grandmother plus grandchildren’). Instead they use the family group
classifier pi55li21 (‘grandfather plus grandchildren’) for groups of any grandparent of
either sex plus grandchildren. These forms are often extended to include spouses in
the descending generations. The three-generation terms can also be used in reference
to the groups spanning up to four generations, thus including grandparents, parents,
children and their children. In case someone from the lower generation is included,
these terms are also employed to refer to groups including more than one person in
the higher generation. This is shown in the range of possible readings in (47).
(47) Lisu (Central Yi, Tibeto-Burman) (Bradley 2001:13)
ku44 pi55 li21
nine NUM.FGCL:GRANDFATHER
‘grandfather plus eight grandchildren’ (or ‘grandfather, grandmother and seven
grandchildren’, or ‘grandfather, grandmother, father, mother and five
grandchildren, or ‘grandfather, father, father’s brother, father’s brother’s wife,
two grandchildren who are children of the father, one of their spouses, their child
(who is a great-grandchild), and a child of the father’s brother’, etc.) or
‘grandmother and eight grandchildren, etc.’

4.3.4. Arguments for the hypothesis


As shown in Section 4.2, the Mayan language Jacaltec, categorizes human referents
according to the kinship criterion (among other criteria). Thus, referents are
categorized as kin vs. non-kin to the speaker. The numeral classifier system in the Yi
languages also has kinship as its main parameter for human categorization. However,
the latter classifier system offers somewhat finer gradations of the categories. Thus, in
the Yi branch languages a referent can be classified not purely according to a kin vs.
non-kin relationship with the speaker or another referent, but also according to the
family relation one is sharing. The examples given above demonstrate that the family
group classifiers of different Yi languages include kin groups like ‘a father plus
children’, ‘a mother plus children’, ‘grandparent plus grandchildren’ (or two
separated, more specific ones, namely ‘grandfather plus grandchildren’ vs.

46
‘grandmother plus grandchildren’). Furthermore, there are family group classifiers
with the meaning ‘brothers’, ‘siblings’, ‘spouses’. Besides, in some of the Yi branch
languages, the core meaning of the family group classifiers can be extended, which
results in an impressive list of possible readings. Bradley (2001:13) mentions that, for
example, the semantics of the numeral classifier in (47), which is used for
grandparents and grandchildren, can be extended to include close family friends. It it
unclear, though, what can warrant such semantic extensions and whether they are
subjective and situational.
On the basis of the data it can be concluded that the choice of an appropriate
numeral classifier in the Yi branch languages highly depends on the kinship relation
of the speaker to his/her referents. Therefore, the numeral classifiers carry information
on the social identities of the speech act participants, namely on the family
relationship between the referents, or between the speaker and a referent. This is
exemplified by the phrase in (46b) and repeated in (48):
55
(48) wa33nu21 sa44 l
we(EXCL) two NUM.FGCL:SIBLING.SPOUSE
‘The three of us’ (siblings or cousins)
The presence of the given qualities of the numeral classifiers in the Yi branch
languages provides arguments in support of the hypothesis that numeral classifiers
can have a socially deictic function.

4.4. Numeral classifiers. Religious address forms: Burmese and Thai


The present sections deals with systems of numeral classifiers which encode
information on spiritual function and status of human referents. As case studies, the
classifier systems of Burmese and Thai, the languages of two highly stratified
societies, are considered.

4.4.1. Burmese
Burmese, a Sino-Tibetan language, is spoken by about 32 million in Myanmar, also
known as Burma. Burmese can be characterized as a numeral classifier language. The
system of numeral classifiers in Burmese falls into two subsystems: one categorizing
inanimate referents, the other one categorizing animates. The latter is of interest for
the present thesis and is considered next.
Becker (1975:115) argues that the system of numeral classifiers used with
animate referents can be presented not hierarchically but in terms of concentric circles
with “Buddha (and his words, relics, and images) as the centre” and all other animate
beings placed “in the network radiating out from the centre.” Table 10 shows the
classifier inventory.
Thus, the position in the centre is devoted to Buddha and the things which are
associated with him, such as his words - the Buddhist Law, relics and images. The
classifier used for it is shu. Becker (1975:116) notes that this classifier can be applied
not only to the whole system itself, such as the Dharma, the field of human existence,
but also by means of metaphorical extension to common objects conceptually similar

47
to the system in shape, such as fishing and mosquito nets, gardens and staircases.
Becker (1975:116) and Foley (1997:237) mention that in traditional Burmese culture
fishing and mosquito nets were conical in shape, gardens were laid out as a wheel,
and staircases were shaped spiralling.

Classifier Semantics
Centre shu Buddha and associated things
st
1 orbit ‘pa (close) deities, saints, monks, royalty
nd
2 orbit ‘u (head) people of status, e.g. teachers, scholars
rd
3 orbit yau  ordinary humans
th
4 orbit kau (body) those lacking wisdom, animals, ghosts, corpses, depraved people, children.

Table 10. Numeral classifiers in Burmese categorizing human beings (adapted from Becker 1975:115,
Foley 1997:238)

On the first ‘orbit’ next to Buddha are all saints, monks, precious things and the
royalty. The classifier used for these beings is ‘pa, the word which also means ‘close’
in modern Burmese.
People not yet close to Buddha but with some spiritual status, such as scholars
and teachers, are placed further away, on the second ‘orbit’. These are classified as ‘u.
This word itself also means ‘head’, ‘beginning’, ‘origin’ and ‘top’.
Next to it, and further away from Buddha, are ordinary people, which are
classified as yau  . This term, unlike several other classifiers, does not refer to
anything else.
Furthest away from Buddha are those lacking wisdom, including animals,
ghosts, depraved people, children and dead bodies. These creatures are classified as
kau , the term also meaning ‘body’. Becker (1975:116) points at the opposition
between head and body which is found a lot in Burmese spiritualism.
The following phrase illustrates the use of classifier yau  with ordinary humans.
(49) Burmese (Sino-Tibetan) (Becker 1975:117)
lu ‘le yau 
person four CLF:PERSON
‘Four people'

4.4.2. Thai
Thai, the national and official language of Thailand, counts about 25 million speakers.
According to Hundius and Kölver (1983:165), Thai is “in the core group of East
Asian numeral classifier languages.” They mention as well that this language in
comparison to other numeral classifier languages of the area has “an extremely
diversified network of separate classes.” Thus, these classifiers form rather a network
than a system, in which three major semantic areas can be framed out. These are: (a)
classifiers for nouns denoting animate beings; (b) classifiers for nouns denoting
inanimate beings; and (c) classifiers for nouns denoting immaterial notions. At the

48
moment, we are interested in the first area, the set of classifiers used for animates.
These are considered next.
Foley (1997:237, citing Hundius and Kölver 1983:193) presents classifiers
used for animate beings along the vertical dimension, running from the classifiers
used for sacred beings to those for profane ones.

sacred
phrá o for the Buddha, deities, royalty
o for the Buddha, deities, royalty, monks (weaker alternate for phrá o )
rûup for priests, monks, idols
thân for persons of high social rank, e.g. teachers, ministers, lesser nobility
naay for men of some social standing
naa for women of some social standing
khon for ordinary persons
ton for beings of supernatural faculties (sinister implication)
chîak for tame elephants
tua for any kind of animal or bird
profane
Figure 6. Numeral classifiers in Thai used for animate beings (adapted from Foley 1997:237)

The first two classifiers phrá o and o are used for superhuman beings. Hundius and
Kölver (1983:193) mention that they are alternatives for each other with the
difference that the former implies yet stronger respect and devotion.
The term o is generally used for royalty, that is the king and queen, princes and
princesses, and dukes and duchesses (Haas 1942:202).
The term rûup is used for priests, monks and idols. According to Hundius and
Kölver (1983:193), the classifier is restricted to very few nouns.
For nobles below the rank of duke and also for persons holding a high official
position, one should use the term thân (Haas 1942:202, Hundius and Kölver
1983:193). This is also the 2nd and 3rd person pronoun which implies respect.
The classifiers naaj and naa are used in reference to, respectively, men and
women of some social standing, public or official position. These are also 2nd person
pronouns which are used with children and younger siblings.
The general classifier for nouns denoting persons, except sacred and royal
ones, is khon which is substitutable for any of the more specific classifiers (Hundius
and Kölver 1983:193). The following example illustrates usage of this classifier.
(50) Thai (Tai) (Hundius and Kölver 1983:167)
kh w mii lûuk s am khon
he have child three CLF:PERSON
'He has three children'
As Haas (1942:202) notes “sacred and awesome beings are also arranged in a series of
ranks which is reflected in the choice of classifier”. Thus, for gods or angels the term
o is generally used. Beings with supernatural faculties which imply sinister aspects,

49
such as hermits, ghosts, spirits, devils, demons, etc., are referred to with the classifier
ton.
The last two terms chîak and tua are used for animals in general. The former is
employed in reference to tamed elephants. The latter, on the other hand, is used for
any kind of animal. Besides, it has another function: the classifier is used to denote
persons acting in some assumed character.
It should also be mentioned that the first six classifiers given in Figure 6,
namely phrá o , o , rûup, thân, naay and naa , can also be used as vocatives
(Crevels, p.c.).

4.4.3. Arguments for the hypothesis


As demonstrated, in Burmese and Thai human referents are categorized mainly
according to their religious and social status. The classifier inventories of both
languages include classifiers which make distinctions between deities, royalty, people
of high or low religious status and function, and ordinary humans. Additionally, in
Thai, male and female referents who have some social standing in the community are
referred to with separate classifiers. In Burmese, on the one hand the gender
parameter is not included into categorization, on the other hand there is a separate
category for children and other referents which are considered to ' lack wisdom' .
The choice of an appropriate classifier is dictated by the social context in
which a speech act takes place. As mentioned earlier, the Thai society is highly
stratified, and this is reflected in every social environment, e.g. at work, on the street,
at home, in school. As Campbell (1998:1) notes, “each new environment that a Thai
person enters calls for a re-assessment of his/her social standing”. Depending on the
social identities of the speech act participants the choice of a classifier will vary.
When used in address or in reference, the spiritual and social function and
status of the addressee or the referent relative to the one of the speaker plays a role.
For example, addressing a male person of a higher social status than the speaker, the
classifier naay is used. Another option is the term thân which is used with persons of
a high social status independent of his/her gender.
The choice of appropriate classifiers can also be dictated by the social identity
of a bystander in a speech event. The presence of a royalty member will require the
use of classifiers, normally not employed in an everyday situation. For instance, the
numeral classifier chîak used for enumerating elephants has much more formal air and
will be used instead of the neutral tua form.
As seen above, the classifier systems in Burmese and Thai are far from being
static. Depending on multiple social factors such as, for instance, the identities of
speech act participants and the social relations between them, the usage of classifiers
will vary. As Becker (1975:115) mentions regarding the categorization in Burmese,
“[p]eople have no fixed position in this network”. Unfortunately, I did not find any
examples illustrating the usage of different classifiers in socially different
environments. Nevertheless, the dynamic properties of the numeral classifier systems
in Burmese and Thai found in the descriptions of these languages can give us ground
to argue that the considered numeral classifiers can adopt a socially deictic function,

50
for they can be anchored to the social identities of the speech act participants.

5. Conclusion and discussion


The present thesis deals with such linguistic phenomena as social deixis and
classifiers used with human referents. The question of social deixis has been
addressed in Lyons (1968), Fillmore (1975) and extensively discussed in Levinson
(1979). As the working definition of the term social deixis I adopt the one proposed
by Levinson (1979:206), namely “those aspects of language structure that are
anchored to the social identities of participants (including bystanders) in the speech
event, or to relations between them, or to relations between them and other referents”.
As Levinson (1979:206) adds, various aspects of language depending on these
relations can be regarded as relevant to social deixis “in so far as they are
grammaticalized”. The classical examples of social deixis are the T/V phenomenon,
honorifics in Japanese and the brother-in-law language in a number of Australian
languages.
Classifiers, on the other hand, as one of the nominal categorization devices
have been studied, among others, by Allan (1977), Aikhenvald (2000), Craig (1986b),
Grinevald (2000), Senft (1996, 2000a,b). In her extensive study on classifiers,
Aikhenvald (2000:260, 284-6, 288, 357) refers to languages which have inventories of
classifiers for humans and discusses the semantics of such classifiers. Day (1973a,b)
and Craig (1986b) present a description of the noun classifier systems in Jacaltec, of
which one is labelled by Craig (1986b) as the system of social interaction. Rehg
(1981) and Keating (1997) describe the set of genitive classifiers in Pohnpeian which
have a status indexing role.
However, there are no studies, as far as I know, which have explicitly
addressed the issue of classifiers and social deixis. Therefore, I have raised the
question in the present thesis and have attempted to test the hypothesis that classifiers
used with human referents can have a socially deictic function.
The question is approached in the following manner: first, the phenomenon of social
deixis is considered and exemplified by several case studies. These case studies
demonstrate how such social parameters as status, kinship, age and sex determine
linguistic behaviour and are reflected in various aspects of the language structure.
Second, the classifiers types are discussed briefly and exemplified. Finally, four
groups of languages and their inventories of classifiers used with human referents are
investigated. The selected groups of languages are as follows:
(a) Pohnpeian;
(b) Jacaltec, Mam and Akateko;
(c) Yi branch languages;
(d) Burmese and Thai.
The choice for the languages is based on several factors, namely:
• in these languages the inventories of classifiers used with human referents are
relatively large in size;

51
• these languages demonstrate that at least numeral, noun and genitive classifier
types can be engaged in the encoding of information on the social identity;
• the classifier inventories of the groups of languages have a particular
dominating social parameter according to which referents are classified.
Namely: social status (genitive classifiers in Pohnpeian); kinship, age and sex
(noun classifiers in Jacaltec), age and sex (noun classifiers in Mam and
Akateko); family membership (numeral classifiers in Yi branch languages);
religious status and function (numeral classifiers in Burmese and Thai).
Within the frame of each language, I focused on the classifier assignment patterns to
see whether, and how, the choice of an appropriate classifier varied depending on the
social identity of the participants in a speech event. It goes without saying that such
task is carried out ideally when a sufficient amount of examples / information is
available. For some of the considered languages (e.g. Mam, Burmese) this kind of
information was very limited.
Nevertheless, it has been observed that the usage of classifiers in the
considered languages depends on the identities and social relations between the
speaker, referent, addressee and, in some cases, a bystander in the speech event. In
other words, classifier forms are frequently anchored to the social identities of speech
act participants, thus meeting the major criterion for socially deictic items in the
definition followed in the thesis. Looking separately at each of the four language
groups, the following observations can be made.
In Pohnpeian, the choice an appropriate genitive classifier is determined by the
rank of addressee or referent relative to the rank of the speaker, and/or the rank of
addressee or referent and speaker relative to the one of a bystander. This is reflected
in the grammatical constructions, some of which show two different levels of status
simultaneously, for instance the status of the speaker and a referent relative to the one
of a bystander.
In Jacaltec, a human referent is also seldom associated with one particular
classifier. Depending on religious and social status, age, sex and kinship relation of
the referent to the speaker the choice of a noun classifier will vary. However, the
choice of a classifier does not always reflect the real social identity of the referent and
his/her relationship to the speaker. Namely, one can ‘play’ with classifiers to create
pragmatic focus and communicate his/her own estimate of the social relationship
between him/her and a referent. For instance, while the standard classifier assignment
scheme dictates the use of the noun classifier naj with the semantic content of ‘male,
non-kin and unmarked for status’, the speaker can either use this classifier or choose
another one out of the noun classifier paradigm available in the language. In the first
case, the classifier usage indexes the actual social identities and relations between
speech act participants. In the second case, the speaker’s subjective estimate of the
referent’s social identity and their social relation is communicated. Mam and Akateko
have inventories of noun classifiers which categorize human referents according to
age, sex and social status. The presence of the classifier forms with semantics
‘respected’ or ‘person of the same status, fondly’, nevertheless, gives us ground to
argue that the usage of at least these classifiers depends on the social relations

52
between speech act participants. For instance, the choice of a classifier is dictated by
the social status of a referent relative to the speaker and/or the estimation of solidarity
between them.
While in Jacaltec human referents are categorized according to kinship
(among other criteria), the numeral classifier system in the Yi branch languages has
kinship, and, to be more precise, family relationship as its main parameter for human
categorization. The semantics of such numeral classifiers can include the kin groups,
like ‘a mother plus children’, ‘grandfather plus grandchildren’, ‘brothers’, ‘siblings’,
‘spouses’, etc. According to Bradley (2001:13), the core meaning of the classifiers
can be extended. For instance, the classifier that is used with grandparents and
grandchildren can be extended to include close family friends. As mentioned earlier, it
is unclear what can warrant such semantic extensions and whether they are subjective
and situational. Nevertheless, the available data show that the choice of a numeral
classifier in the Yi branch languages highly depends on the kinship relation of the
speaker and/or his/her referents. Thus, the numeral classifiers are able to carry
information on the social identities of the speech act participants, namely on the
family relationship between the referents, or between the speaker and a referent(s).
In Thai and Burmese, humans are categorized mainly according to their
religious and social status and function. The numeral classifiers in the languages offer
categorization of human referents into deities, royalty, monks, priests, people of high
social status, people of some social standing and ordinary humans. Besides, there is a
classifier for the Buddha and the things associated with him, and a classifier for
profane beings, such as ghosts and those “lacking wisdom”. For Burmese, no concrete
examples are available regarding standard classifier assignment schemes, although
Becker (1975:115) mentions that “[p]eople have no fixed position in this network”.
He also gives an interesting and sharp characteristic of the classifier system saying
that “if one considered a king to be depraved, he might classify him in private as an
animal, though it might be wise and safe to classify him in public as a saint”. Such
capacity for reclassification, even for pragmatic purposes, gives us reasons to believe
that the numeral classifier system in Burmese is rather dynamic. The presence of
separate categories for ordinary humans, high status individuals, royalty and monks,
to name just a few, presumes that the categorization process is able to reflect the
social identity of a referent. Besides, in most cases categorization, or the statement
about the absolute position of the referent, has to be made from the ‘relative’
perspective of the speaker. This is also the case in Thai. In this language, many of the
numeral classifiers can also be used as vocatives, the feature that gives some extra
room for verbal play. However, when either used in address or in reference, the
spiritual or social status of the addressee, or the referent, relative to the one of the
speaker determines the choice of a classifier. As in Pohnpeian, in Thai (for the other
languages I have no direct information), the choice of appropriate classifiers can also
be dictated by the social identity of a bystander in a speech event. This takes form not
only in the lexical alternatives, such as honorifically marked nouns, but also in the
change of classifiers.
It can be noticed that the ‘core’ classifier which is present in each of the

53
considered classifier inventories is the one marking status. For the present set of
languages the following implication is valid: if a language has a classifier which
marks status, it has at least one classifier unmarked for status, with the semantics
‘ordinary’.
However, the phenomenon of social deixis, in the way it is defined in the thesis, will
not be observed when a classifier language provides a division of nouns (a) only
according to humanness or animacy; and/or (b) according to the parameter of sex
(male vs. female). Those classifiers which make reference to humanness, or sex of a
referent noun, do not contain or reflect information on the social identity of any other
speech act participant. For instance, the Mam and Akateko languages, beside the
classifiers categorizing referents according to status/respect, employ classifiers which
categorize humans exclusively by sex. The choice and use of the latter do not require
a relative standpoint and are not determined by the social identities of other speech act
participants, for instance the one of the speaker. The choice is determined by sex of
the referent only. Therefore, it can be speculated that in order to qualify as a socially
deictic item (again the followed definition is important), classifiers used with humans
should make reference to at least one of the following parameters: status, spiritual
status or function, and kinship, be it in combination with reference to sex and/or age,
or without it. This is motivated by the fact that assignment of the classifiers which
include either of these parameters usually requires a relative standpoint. Classifiers
categorizing referents according to the parameter of age, probably can refer either to
the ‘absolute’ age-group to which the referent belongs or to the age-group relative to
the one of the speaker or another referent in the speech event. In the latter case, the
classifier choice and use would show variation depending on the social identities of
the speech act participants. However, this is a debatable question and remains to be
studied further.
As to the other criterion for social deixis, namely to constitute a
grammaticalized aspect of a language, the following points can be mentioned. In
Section 3 it has been noted that classifiers differ by their qualities from such purely
lexical classification systems as measure terms and class terms, yet they are distinct
from the purely grammatical ones such as noun class and gender. The intermediate,
lexico-grammatical, position of classifiers accounts for the overlap or blending
between the systems, which results in borrowing of characteristic features of either
the lexical or the grammatical end. It must be mentioned at this point that the term
‘classifier’, used throughout the analysis in reference to the classifying morphemes in
the considered languages, has been adopted from original sources and descriptions of
the given languages. Therefore, it is not implausible that in the various sources
slightly different criteria have been used for the term ‘classifier’. Among the systems
considered in the thesis, the Jacaltec system of noun classifiers described in Craig
(1986b) can be argued to have strong features of class terms (Enfield, p.c.), thus
sooner representing a lexical than a grammatical classification device. In this case, the
given noun classifiers in Jacaltec may not satisfy the grammaticalization criterion for
socially deictic items. Still they satisfy the criterion of being able to index the social
identities of speech act participants and reflect information on social relations

54
between them.
Concluding it is noted that the results of the analysis of the present data argue
in support of the hypothesis that classifiers used for human categorization have a
socially deictic function, provided that the categorization includes the parameters
outlined above, namely: social status, religious status and function, and kinship, be it
in combination with categorization according to sex and/or age, or without it.
Regarding prospects for a future study, the following point can be mentioned. The
testing of the hypothesis that classifiers used with human referents can have a socially
deictic function has been approached in the present thesis from one perspective.
Namely, whether and how the choice of appropriate classifiers varies parallel to
alternations in the social situation in which the speech act occurs. The question can
also be addressed from a different angle. In Section 2.2.4, the observations made by
Inoue (1979) are mentioned regarding the usage of honorifics in modern Japanese.
Inoue (1979:295) argues that the deferential language may sooner be a marker of the
social identity of a speaker, since in modern times knowledge of deferential
expressions is a feature of linguistic prestige. In other words, the ability to use such
expressions appropriately is considered to be a mark of good upbringing. According
to Inoue (1979:296), the fact that well-educated intellectuals tend to speak in a more
polite language regardless of who they are speaking to, can be interpreted rather as a
reflection of the estimate of their own social position than a reflection of the respect
of others. Such tendency has been observed with use of classifiers in Lao (Enfield,
p.c.). Similarly, Carpenter (1986:23) in her work on Thai classifiers mentions a
running gag in one TV situation comedy including a peasant servant-girl from a rural
area who overused the general classifier nooey from her dialect, thus producing
sentences like mii khaek maa s ng nooey ' Two guest-things have arrived' . As
Carpenter (1986:23) notes “the source of the humor was more her uneducated
“country-bumpkin” style of speech than the misclassification per se”, thus serving as
an example of the classifier usage reflecting the social identity of the speaker. This
interesting phenomenon has not been addressed in the present study for the following
reason: we would hardly observe influence of a speech act situation environment on
the choice and use of classifiers. In other words, the choice and use of classifiers
would depend on and reflect the social identity of the speaker or the estimate of
his/her own identity, barely encoding any information on the social identity of other
speech act participants. Thus, the latter approach to the issue of social deixis and
classifiers can constitute a research question for a further study. In that case, it is
advisable not to limit the research data to the classifiers used with humans, but to
expand the focus by considering the entire classifier paradigm present in a given
language.

55
References
Adams, K. (1986) Numeral classifiers in Austroasiatic, in C.G. Craig (ed.), pp.241-
262.
Adelaar W.F.H. with the collaboration of P.C. Muysken (2004) The Languages of the
Andes. Cambridge: Cambridge University Press.
Aikhenvald, A.Y. (2000) Classifiers: A Typology of Noun Categorization Devices.
Oxford: Oxford University Press.
Allan, K. (1977) Classifiers. Language 53(2):285-311.
Alpatov, V.M. (2006) Kategorija Vezhlivosti v Sovremennom Japonskom Jazyke
[Category of Politeness in Modern Japanese]. Moscow: DomKniga.
Altman, G. and A. Riska (1966) Towards a typology of courtesy in language.
Anthropological Linguistics 8:1-10.
Ayu, T. (1989) Liangshan Yiyu liangci [Liangshan Yi classifiers]. Chengdu, Sichuan:
Xinan Minzuxueyuan, Minzu Yuyanwenxueyanjiu suo [Southwest Institute of
Nationalities, Minority Language and Literature Research Institute].
Barz, R.K. and A.V.N. Diller (1985) Classifiers and standardisation: some South and
South-East Asian comparisons, in D. Bradley (ed.) Language Policy, Language
Planning and Sociolinguistics in South-East Asia. Canberra: Pacific Linguistics, pp.
155-84.
Becker, A.L. (1975) A linguistic image of nature: Burmese numeral classifiers.
Linguistics 165:109-121.
Benton, R.A. (1968) Numeral and attributive classifiers in Trukese. Oceanic
Linguistics 7:104-146.
Bradley, D. (2001) Counting the family: family group classifiers in Yi (Tibeto-
Burman) languages. Anthropological Linguistics 43(1):1-17
Brown, R. and A. Gilman (1972) The pronouns of power and solidarity, in P.P.
Giglioli (ed.) Language and Social Context. London: Penguin Books, pp. 252-282.
Cambell, L. (1997) American Indian Languages. New York: Oxford University Press.
Campbell, S. (1998) Personal pronoun usage in Thai. Available at:
www.scampnet.net/Personal-Website/thai_files/thai_pronouns.pdf
Carpenter, K. (1986) Productivity and pragmatics of Thai classifiers. Proceedings of
the Twelfth Annual Meeting of the Berkeley Linguistic Society. Berkeley, CA:
Berkeley Linguistics Society, pp. 14-25.
Collins, B. (1962) Tonga Grammar. London: Longmans.
Craig, C.G. (1986a) Introduction, in C.G. Craig (ed.), pp.3-12.
----- (1986b) Jacaltec noun classifiers: a study in language and culture, in C.G. Graig
(ed.), pp. 263-293.
----- (1990a) Chibchan nominal classification. Paper read at the annual meeting of
the American Anthropological Association.

56
----- (1990b) Rama Kuup: gramática rama. Managua: Centro de investigaciones y
documentación de la Costa Atlántica.
----- (2000) Classifiers, in G. Booij, C. Lehmann & J. Mugdan (eds.) Morphology:
An International Handbook on Inflection and Word-Formation. Berlin: Mouton de
Gruyter.
----- (ed.) (1986) Noun Classes and Categorization. Amsterdam: John Benjamins.
Crevels, M. (2003) Parts of speech in Itonama. Handout for Americanist Seminar,
Nijmegen, April 14, 2003.
Day, C. (1967) The Jacaltec language. Unpublished doctoral dissertation, Department
of Anthropology, University of Chicago.
----- (1973a) The Jacaltec Language. Bloomington: Indiana University Publications.
----- (1973b) The semantics of social categories in a transformational grammar of
Jacaltec, in M.S. Edmonson (ed.) Meaning in Mayan Languages. The Hague:
Mouton, pp. 85-105.
DeLancey, S. (1986) Toward a history of Tai classifier systems, in C.G. Graig (ed.),
pp. 437-452.
Denny, J.P. (1976) What are noun classifiers good for? Papers from the Twelfth
Regional Meeting of the Chicago Linguistic Society, pp.122-132.
Dixon, R.M.W. (1982) Noun classifiers and noun classes, in R.M.W. Dixon Where
Have All the Adjectives Gone? And Other Essays on Semantics and Syntax. The
Hague: Mouton, pp. 211-233.
----- (1986) Noun classes and noun classification in typological perspective, in C.G.
Craig (ed.), pp. 105-112.
Dixon, R.M.W. and A.Y. Aikhenvald (eds.) (1999) The Amazonian Languages.
Cambridge: Cambridge University Press.
Duranti, A. (1992) Language in context and language as context: the Samoan respect
vocabulary, in A. Duranti and C. Goodwins (eds.) Rethinking Context: Language as
an Interactive Phenomenon. Cambridge: Cambridge University Press.
Durie, M. (1985) A Grammar of Acehnese: On the Basis of a Dialect of North Aceh.
Dordrecht: Foris.
England, N.C. (1983) A Grammar of Mam, a Mayan language. Austin: University of
Texas Press.
Erbaugh, M. (1986) Taking stock: the development of Chinese noun classifiers
historically and in young children, in C.G. Craig (ed.), pp. 399-436.
Evans, N. (1996) The syntax and semantics of body part incorporation in Mayali, in
H. Chappell and W. McGregor (eds.) The Grammar of Inalienability: A Typological
Perspective on Body Part Terms and the Part-Whole Relation. Berlin: Mouton de
Gruyter, pp. 65-109.
Fillmore, C.J. (1975) Santa Cruz Lectures on Deixis, 1971. Bloomington: Indiana
University Linguistic Club.

57
Fischer, J.L. (1969) Honorific speech and social structure: a comparison of Japanese
and Ponapean. Journal of Polynesian Society 78(3):417-422.
Foley, W. (1997) Anthropological Linguistics: An Introduction. Oxford: Basil
Blackwell.
Fried, M. (1967) The Evolution of Political Society. New York: Random House.
Friedrich, P. (1966) Structural implications of Russian pronominal usage, in W.
Bright (ed.) Sociolinguistics. The Hague: Mouton.
Garvin, P.L. and S.H. Riesenberg (1952) Respect behavior on Ponape: an
ethnolinguistic study. American Anthropologist 54:201-220.
Giddens, A. (1987), (3rd edn 1997) Sociology. Cambridge: Polity.
Grinevald, C. (2000) A morphosyntactic typology of classifiers, in G. Senft (ed.), pp.
50-88.
Haas, M. R. (1942) The use of numeral classifiers in Thai. Language 18(1):201-205.
Hatsutaro, O. (1968) Tadashii Keigo. (Reference found in Alpatov 2006).
Haviland, J. (1979) How to talk to your brother-in-law in Guugu-Yimidhirr, in T.
Shopen (ed.), pp. 161-239.
Headley, R.K, Jr., K. Chhor, L.K. Lin, L.H. Kheang and C. Chun (1977) Cambodian-
English Dictionary. 2 Volumes. Washington, D.C.: Catholic University of American
Press.
Hundius, H. and U. Kölver (1983) Syntax and semantics of numeral classifiers in
Thai. Studies in Language 7(2):165-214.
Inoue, K. (1979) Japanese: a story of a language and people, in T. Shopen (ed.), pp.
241-300.
Jensen, J.T. (1977) Yapese Reference Grammar. Honolulu: University Press of
Hawaii.
Keating, E. (1997) Honorific possession: power and language in Pohnpei, Micronesia.
Language in Society 26:247-268.
Lakoff, G. (1986) Classifiers as a reflection of mind, in Craig (ed.), pp. 13-51.
Levinson, S.C. (1979) Pragmatics and social deixis, in C. Chiarello (ed.) Proceedings
of the Fifth Annual Meeting of the Berkeley Linguistic Society. Berkeley, CA:
Berkeley Linguistics Society.
Lichtenberk, F. (1983) Relational classifiers. Lingua 60:147-176.
Lowe, I. (1999) Nambiquara, in R.M.W. Dixon and A.Y. Aikhenvald (eds.), pp. 269-
291.
Lyons, J. (1968) Introduction to Theoretical Linguistics. Cambridge: Cambridge
University Press.
Manning, P. (2001) On social deixis. Anthropological Linguistics 43(1):54-100.

Marnita, R. (1996) Classifiers in Minangkabau. MA thesis, Australian National


University, Canberra.

58
Martins, S. (1994) Análise da morfosintaxe da língua Dâw (Maku-Kamã) e sua
classificação tipológica. MA thesis, Universidade Federal de Santa Catarina,
Florianópolis, Brazil.
Martins, S. and V. Martins (1999) Makú, in R.M.W. Dixon and A.Y. Aikhenvald
(eds.), pp. 251-267.
Maspero, G. (1915) Grammaire de la Langue Khmèr (Cambodgien). Paris.
Mateo Toledo, E. (1999) La cuestión Akateko-Q’anjob’al. Una comparación
gramatical. MA-thesis. Universidad Mariano Galvez de Guatemala.
Matsumoto, Y. (1993) Japanese numeral classifiers: a study of semantic categories
and lexical organization. Linguistics 31:667-713.
Mattéi-Müller, M.C. (1976) Cuestiones de categorización en las construcciones
posesivas del panare. Antropológica 38:3-14.
Mesthrie, R., J. Swann, A. Deumert and W. Leap (2000) Introducing Sociolinguistics.
Edinburgh: Edinburgh University Press.
Moore, B. and G. Franklin (1979) Breves Notícias da Língua Makú-Hupda. Brasília:
Summer Institute of Linguistics.
Petersen, G. (1982) One Man Cannot Rule a Thousand. Ann Arbor: University of
Michigan Press.
Pinsker, E. (1997) Traditional leaders today in the Federated States of Micronesia, in
G.M. White and L. Lindstrom (eds.) Chiefs Today: Traditional Pacific Leadership
and the Postcolonial State. Stanford, CA: Stanford University Press.
Rehg, K. (with D.C. Sohl) (1981) Ponapean Reference Grammar. Honolulu:
University of Hawaii Press.
Rodrigues, A.D. (1995) Some morphological and syntactic aspects of Karirí. Paper
presented at the SSILA meeting, Albuquerque, NM.
Sánchez, M. (1977) Language acquisition and language change: Japanese numeral
classifiers, in G. Blount and M. Sánchez (eds.) Sociocultural Dimensions of Language
Change. New-York: Academic Press, pp. 51-62.
Senft, G. (1996) Classificatory Particles in Kilivila. New-York: Oxford University
Press.
----- (2000) What do we really know about nominal classification systems? in G.
Senft (ed.), pp. 11-49.
----- (ed.) (2000) Systems of Nominal Classification. Cambridge: Cambridge
University Press.
Sherzer, J. (1976) An Areal-Typological Study of American Indian Languages North
of Mexico. Amsterdam: North-Holland.
Shibatani, M. (1990) The Languages of Japan. Cambridge: Cambridge University
Press.
Shopen, T. (ed.) (1979) Languages and their Speakers. Cambridge, MA: Winthrop

59
Tilakaratne, S. (1988) Social deixis in Sinhalese: the pronoun system. Kansas
Working Papers in Linguistics 13:174-190.
Thompson, D. (1995) The Concise Oxford Dictionary. Ninth edition. Oxford:
Clarendon Press.
Uchida, N. and M. Imai (1996) A study on the acquisition of numeral classifiers
among young children: the development of human-animal categories and
generalization of the rule of classifier applying. Japanese Journal of Educational
Psychology 44:126-135.
Yamamoto, K. (2005) The Acquisition of Numeral Classifiers: The Case of Japanese
Children. Berlin: Mouton de Gruyter.

60

You might also like