Professional Documents
Culture Documents
Using Japanese Honorific Expressions: A Psychological Study: Psychology
Using Japanese Honorific Expressions: A Psychological Study: Psychology
A Psychological Study
TAMOTSU SHIRADO
National Institute of Information and Communications Technology
(NICT), Kyoto
SATOKO MARUMOTO
The Institute of Behavioral Sciences (IBS), Tokyo
MASAKI MURATA AND HITOSHI ISAHARA
National Institute of Information and Communications Technology
(NICT), Kyoto
________________________________________________________________________
We investigated, via experiment, knowledge of normative honorific expressions as used in textbooks and in
practice by people. Forty subjects divided into four groups according to age (younger/older) and gender
(male/female) participated in the experiments. The results show that knowledge about the use of normative
honorific expressions in textbooks is similar to that demonstrated by the younger subject groups, but differed
from that of the older subject groups. The knowledge of the older subjects was more complex than that shown
in textbooks or demonstrated by the younger subjects. A model that can identify misuse of honorific
expressions in sentences is the framework for this investigation. The model is minimal, but could represent 76%
to 92% of the subjects knowledge regarding each honorific element. This model will be useful in the
development of computer-aided systems to help teach how honorific expressions should be used.
Categories and Subject Descriptors: J.4 [Computer Applications]: Social and Behavioral Sciences--
Psychology
General Terms: Human Factors, Experimentation, Measurement, Performance
Additional Key Words and Phrases: Japanese, honorific expressions, misuse
________________________________________________________________________
1. INTRODUCTION
In Japan, politeness plays an important role in social activities, especially in conversation.
There are few languages in which politeness is as highly developed as it is in Japanese,
[Hayashi et al. 1974; Suzuki et al. 1984]; hence the correct use of Japanese honorific
expressions is essential to facilitate smooth communication. In fact, most entrance
examinations and entry-level training programs at Japanese companies include the correct
use of these expressions (and education on this subject has been increasing).
Although many textbooks on Japanese honorific expressions have been published
recently, they are not necessarily consistent with what many people know about
normative honorific expressions. Most of the textbooks are based on traditional studies,
but linguistic norms have changed over time. So to understand how normative honorific
expressions can be useful for practical conversations, consistency between textbook and
practical knowledge must be investigated.
________________________________________________________________________________
Authors' addresses: T. Shirado, M. Murata, and H. Isahara, National Institute of Information and
Communications Technology (NICT), Kyoto 619-0289, Japan; S. Marumoto, The Institute of Behavioral
Sciences, Tokyo 162-0845, Japan.
Permission to make digital/hard copy of part of this work for personal or classroom use is granted without fee
provided that the copies are not made or distributed for profit or commercial advantage, the copyright notice,
the title of the publication, and its date of appear, and notice is given that copying is by permission of the ACM,
Inc. To copy otherwise, to republish, to post on servers, or to redistribute to lists, requires prior specific
permission and/or a fee. Permission may be requested from the Publications Dept., ACM, Inc., 2 Penn Plaza,
New York, NY 10121, USA, fax:+1(212) 869-0481, permissions@acm.org
2006 ACM 1073-0516/06/0600-0146 5.00
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006. Pages 146-164.
Using Japanese Honorific Expressions: A Psychological Study 147
The main purpose of our study is to examine use of normative honorific expressions in
textbooks and to compare it to the way people use them. First, we reviewed the types of
honorific expressions and the categories of misuse, keeping in mind that "misuse" is a
relative concept. That is, the word "misuse" simply represents a usage of an expression
that is not consistent with certain linguistic norms, which may change over time, as
mentioned above. Based on these reviews, we introduced as the framework of our study a
model to identify the misuse of honorific expressions in a sentence. We then, via
psychological experiments, examined the similarities and differences in the knowledge
provided in textbooks and by younger and older people. Our model will be useful in the
development of computer-aided systems to help teach how honorific expressions should
be used.
2. MISUSE OF HONORIFIC EXPRESSIONS
2.1 Types of Honorific Expressions
The honorific expressions commonly used in spoken Japanese can be divided into three
types. 1
Type 1: Subject honorific expressions. (Sonkeigo) used to show respect to a person
who is usually the subject of a predicate in the sentence by elevating the status of the
person. Subject honorific expressions principally include (1) honorific titles such as
san/sama, etc; (2) respectful forms of verbs; (3) verb stems2 + auxiliary verbs; and (4)
o/go (prefixes) + verb stems + auxiliary verbs. The auxiliary verb nasaru can be used in
both (3) and (4); the auxiliary verbs sareru/tekudasaru/reru/rareru can be used in (3)
only; and ninaru/kudasaru can be used in (4) only. The honorific title, kun, is not
included because kun is an exception in that it is an honorific title that does not indicate
politeness.
Type 2: Object honorific expressions. (Kenjougo) used to show respect to a person
who is usually the object of a predicate in the sentence by humbling the speaker or the
subject of the predicate. Object honorific expressions principally include (1) humble
forms of verbs; (2) verb stems + auxiliary verbs; and (3) o/go (prefixes) + verb stems +
auxiliary verbs. The auxiliary verb itasu can be used in both (2) and (3); the auxiliary
verb teitadaku can be used in (2) only; and suru/itadaku/mousiageru can be used in (3)
only.
Type 3: Polite expressions. (Teineigo) used to show politeness, but not necessarily
respect, toward a person who is usually the listener. Polite expressions include an
auxiliary verb such as desu/masu/gozaimasu at the end of a sentence.
Different types of expressions can exist in the same sentence. The predicate, especially,
can be both a subject honorific expression and an object honorific expression, called a
"two-directional honorific expression."
2.2 Categories of Misuse
The misuse of honorific expressions can be divided into two principal categories.
Category 1. Misuse of word forms. The word form is not normative; i.e., the word
form differs from the normative word forms mentioned in Section 2.1. For example, o (a
prefix) + verb stem + rareru (an auxiliary verb) is a misuse of word form because rareru
is not allowed in this word form. In this case, the correct expression is o + verb stem +
ninaru.
1
Some vowels had be interpolated for some expressions for types 1 and 2 -- for example, the vowel i had be
interpolated between o (prefix)+kik (stem), and ninaru (an auxiliary verb).
2
The stem is defined as the portion of the verb that is not conjugated.
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
148 T. Shirado et al.
Category 2. Misuse in performance. The word form is normative, but is not used
consistently due to the various social (hierarchical) positions of the speakers, listeners,
and others being addressed or referred to in the sentence. For example, a textbook on
Japanese honorific expressions will generally describe a sentence that does not contain
any polite expressions (i.e., a type 3 expression, see Section 2.1) as a misuse in
performance when the social position of the listener is higher than that of the speaker. We
may think that other factors such as social distance (e.g., familiarity) between persons and
the degree of loss of social prestige that the speech poses to the other person may also
affect the choice of honorific expressions [Goffman 1967; Grundy 1995]. Actually,
however, most textbooks describe the relative social positions of the participants as the
only, or the principal, factor related to serious misuse in performance. Thus to avoid
misuse in performance, social position is the most important factor to take into account,
so we took the relative social positions of the participants as the only factor related to
category 2 mistakes.
Although the misuse of word forms (category 1) is considered as serious as category 2
errors, those in category 1 involve simple mistakes in speech patterns. Thus it is possible
to avoid making category 1 mistakes by simply preparing a list of normative and/or non-
normative word forms. As we are interested in misuse related to speech situations, the
present study concentrates on category 2 problems only.
3. FRAMEWORK
As the framework for our study we introduced a model that can identify the misuse of
honorific expressions in a given sentence.
3.1 Restrictions
The model deals with sentences that satisfy the following restrictions:
(1) Four people a speaker, a listener, and two other persons (referred to as "A" and "B")
are involved.
(2) Only one predicate, whose subject and object are person A and person B, respectively,
is included.
The first restriction is not very strict, since five or more persons are rarely involved in a
sentence, except in extremely complicated conversations. The model requires that we
make the assumption that exactly four people are involved in a sentence, but if it is
slightly modified it can also deal with less complicated situations (i.e., where person A
and/or person B are not involved). The second restriction requires that sentences be
simple so that we do not have to incorporate a parsing program into the model because as
yet there is no parsing program that can accurately identify the subject and object persons
of each predicate in complicated sentences.
Input:
Sentence: "A san ga B san ni osshatta sou da."
Position label: ABLS
Fig. 1. An example of model input and output.
Output:
Decision: "misuse"
Portion of misuse: "da"
Kind of misuse: "impolite"
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
Using Japanese Honorific Expressions: A Psychological Study 149
3
The difference in honorific levels between expressions is not taken into account for value assignment; ex.
E=1 (i.e., honorific type of auxiliary verb at the end of the sentence is polite) is assigned to both gozaimasu
and masu, while gozaimasu is much more polite than masu.
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
150 T. Shirado et al.
row of morphemes
honorific pattern
Output
Fig. 2. The models process flowchart.
is "1 1 0 1" because there are honorific titles (san) for person A and person B, the
honorific type of auxiliary verb (da) at the end of the sentence is impolite, and the
honorific type of the predicate (osshatta) is respectful but not humble.
3.4 Algorithm
Figure 2 shows the flowchart of the model.
The process consists of the following steps:
Step 1. Obtain a row of morphemes from the input sentence by using the ChaSen
Japanese morphological analysis program [Matsumoto 1999].
Step 2. Check for Category 1 misuse (misuse of word forms) in the partial row of
morphemes by using the honorific dictionary shown in Table II (lower portion of the
table). If any misuse is found, proceed to Step 2-1; otherwise, proceed to Step 2-2.
Step 2-1. Output "misuse of word forms," and all the portions of the input sentence
corresponding to the partial rows that have misuses in word form. Then quit the process.
Step 2-2. Check the honorific type of each partial row of morphemes by using the
honorific dictionary (upper portion of Table II), and then determine the values of
honorific elements A, B, E, and P (this organizes the honorific pattern) by using the value
assignment rule shown in Table I.
Step 3. Check the consistency of the honorific pattern with the input position label by
using the consistency check rule shown as Figure 3. This rule refers to the consistency
table, which defines the consistency between the honorific pattern for an input sentence
and the input position label. Details of the consistency table are explained in the next
section.
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
Using Japanese Honorific Expressions: A Psychological Study 151
IF
The honorific pattern obtained for the input sentence
is consistent with the input position label
THEN
Decision: "input sentence is correct."
ELSE
Decision: "misuse."
Portion of misuse: portions of the input sentence corresponding to
honorific elements whose values are not consistent with the position label.
Kind of misuse: honorific types of honorific elements whose values are not
consistent with the position label.
Step 4. If the honorific pattern is defined as consistent with the position label in the
consistency table, output "the input sentence is correct"; otherwise, output "misuse" i.e.,
the portions of the input sentence corresponding to honorific elements whose values are
not consistent with the position label and the honorific types of these honorific elements.
3.5 Consistency Table from Textbooks
Table III shows the consistency table, referred to by a consistency check rule, as
described in the previous section. Each line of the table describes each of the position
labels (24 variations in total, as mentioned in Section 3.2) and honorific patterns that are
consistent with the position label. Only one honorific pattern exists for each of the
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
152 T. Shirado et al.
position labels (except LBAS, BALS, BLAS, and BASL, each of which has two
alternative honorific patterns where the values of the honorific element P differ among
the alternatives).
Table III was obtained from textbooks on Japanese honorific expressions. We assumed
12 speech intentions (Table IV) to obtain and evaluate Table III. Speech intentions (1) to
(6) were used to obtain Table III, and speech intentions (7) to (12) to evaluate it. We
prepared several variations of speech intentions so that we could investigate whether the
consistency table depends on these variations; all speech intentions in Table IV were
adopted based on two criteria:
For each of the speech intentions (1) to (6) in Table IV, we obtained a consistency
table through the following procedure:
Step 1. More than three training sets were prepared for each kind of position label
(24 variations in total). Each training set consisted of the position label and a
sentence. The sentence satisfied the restrictions stated in Section 3.1, and was
considered a normative sentence correspondin to the speech intention under the
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
Using Japanese Honorific Expressions: A Psychological Study 153
relative social position of the four persons specified by the position label. Some of the
sentences were extracted from textbooks on Japanese honorific expressions [Hagino
2001; Hoshino et al. 1993; Horikawa et al. 1969; Kikuchi 1996; Kikuchi 1997; Matsuoka
et al. 1989; Moriyama 2000; Mizutani 1995; Watanabe 1971]. Because it was difficult to
find sentences corresponding to some variations of the position labels in the textbooks,
we composed the other sentences while trying to follow the textbook examples as much
as possible.4
Step 2. For each kind of position label, all kinds of honorific patterns that appeared
together with the position label in the training sets were listed.
Step 3. All the lists obtained in step 2 were tabulated so that each position label and the
honorific patterns corresponding to the position label are described on the same line of
the table.
All six of the consistency tables (one for each of the speech intentions in (1) to (6) in
Table IV) are the same as in Table III. These results suggest that the consistency table
was not affected by variations in the speech intentions in the experiments.
3.6 Validity Check of the Consistency Table Using Textbooks
We verified Table III using two kinds of test sets.
The first kind of test set did not contain any examples of misuse. The second kind
contained examples of misuse in performance (i.e., category 2). In the latter case, we
4
Preliminary experiments showed that there is no difference between the honorific patterns in sentences
extracted from the textbooks and those we composed when both the speech intention and the position label are
the same in each. Hence the method used to prepare the training sets is valid.
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
154 T. Shirado et al.
assumed no misuses in word form (i.e., category 1), since they would not affect the
consistency table as described in Sections 3.4 and 3.5. More than one test set of each kind,
each with the same format as the training set explained in Section 3.5, was prepared for
each of the position labels (24 variations in total) for each of the speech intentions -- (7)
to (12) in Table IV-- via a procedure similar to one used to prepare the training sets.
The above test sets were given to the model one by one. The model did not erroneously
point out any misuse in the first kind of test set, but correctly pointed out all the portions
and kinds of misuse in test sets of the other kind.
These results suggest that Table III correctly represents the knowledge provided by
textbooks regarding Japanese honorific expressions.
3.7 The Summarized Consistency Table from Textbooks
Table III is actually redundant because the relative social positions of the speaker, the
listener, person A, and person B are partially related to the limited honorific elements.
Table III can be summarized as Table V (the summarized consistency table). Table V
was obtained by executing the following steps for each of the honorific elements A, B, E,
and P. The values of honorific elements A, B, and E are each 0/1, and the values of the
honorific element P are 0/1/2/{1 or 3}, where P={1 or 3} corresponds to the position
labels LBAS, BALS, BLAS, and BASL and P=1 does not correspond to these position
labels in step 1.
Step 1. For each value of an honorific element, gather into one group all the position
labels corresponding to the honorific patterns that include the values of the honorific
elements in Table III.
Step 2. For each group obtained in step 1 find the partial relative position commonly
observed in all the position labels of the group, but not observed in the position labels of
other groups at all, where the "partial relative position" represents the order among two or
three symbols in the position label. Specifically, the partial relative position is the logical
expression of X>Y, where X>Y means that symbol X appears to the left of symbol Y in
the position label (X, Y= S, L, A or B, where X differs from Y). So X>Y means that the
social position of the person corresponding to symbol X is higher than that corresponding
to symbol Y. We used a method similar to Karnaugh mapping [Karnaugh 1953] to find
the partial relative position.
For example, the part of Table V for honorific element A was obtained as follows. In
Step 1, the group of position labels corresponding to A=0 and that corresponding to A=1
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
Using Japanese Honorific Expressions: A Psychological Study 155
were obtained. The former group was composed of SLAB, SLBA, SALB, SABL, SBLA,
SBAL, LSAB, LSBA, LBSA, BLSA, BSAL, and BSLA, and the latter group was
composed of the position labels LASB, LABS, ABLS, ALBS, LBAS, BALS, BLAS,
ABSL, ASBL, ASLB, ALSB, and BASL. In Step 2, S>A was obtained as the partial
relative position corresponding to A=0 because S>A is commonly observed in all the
position labels of the former group, but is not observed in the position labels of the latter
group at all. In contrast, A>S was obtained as the partial relative position that
corresponds to A=1 because A>S is commonly observed in all the position labels of the
latter group, but not in the position labels of the former group at all. The parts of Table V
for honorific elements B and E were obtained in a similar way.
The data in Table V suggests the following:
(1) An honorific title does not follow A (A indicates a person's name; (A=0) when the
social position of person A is lower than that of the speaker (S>A). An honorific
title follows A; (A=1) when the social position of person A is higher than that of
the speaker (A>S).
(2) An honorific title does not follow B (B indicates a person's name); (B=0) when the
social position of person B is lower than that of the speaker (S>B). An honorific
title follows B; (B=1) when the social position of person B is higher than that of
the speaker (B>S).
(3) The honorific type of the auxiliary verb at the end of a sentence is impolite, or
there is no auxiliary verb at the end of a sentence (E=0) when the social position
of the listener is lower than that of the speaker (S>L). The honorific type of the
auxiliary verb at the end of a sentence is polite (E=1) when the social position of
the listener is higher than that of the speaker (L>S).
(4) The predicate is neither respectful nor humble (P=0) when the social position of
the speaker is higher than those of person A; person B (S>A S>B), is respectful
but not humble (P=1) when the social position of person A is higher than those of
person B and the speaker (A>B A>S); it is humble but not respectful (P=2) when
the social position of person B is higher than that of the speaker and the social
position of the speaker is higher than that of person A (B>S>A).
(5) Both the respectful and the two-directional honorific expressions (P=1 or 3) can be
used as the predicate when the social position of person B is higher than that of
person A and the social position of person A is higher than that of the speaker
(B>A>S).
4. PSYCHOLOGICAL EXPERIMENT
We performed psychological experiments to compare knowledge of Japanese honorific
expressions as provided by textbooks with knowledge demonstrated by people.
4.1 Procedure
Forty subjects divided into four groups according to age (Y: younger, 18 to 22 years old,
all university students; O: older, over 35 years old; and gender (M, male and F, female)
participated in the experiments. Each group was made up of 10 subjects; we refer to the
groups as YM, YF, OM, and OF, respectively.
The speech intentions shown as Table IV, which were used to obtain and evaluate
Table III, were also assumed in the experiments. For each combination of speech
intentions (12 variations) and the position labels (24 variations), each subject was
required to describe more than one sentence under the following conditions.
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
156 T. Shirado et al.
Therefore, each subject was required to write at least 288 sentences (i.e., 12 speech
intentions 24 position label variations).
4.2 Ratios of Misuse of Word Forms
As explained in Section 2.2, the misuse of honorific expressions can be divided into two
main areas: misuse of word forms (category 1), and misuse in performance (category 2).
To investigate the tendency of subjects to make mistakes in category 1 usage, we
calculated the ratios of category 1 misuse for each subject group. This ratio was defined
as the number of answers that include misuses in word form, divided by the total number
of answers from the subject group.
The results show that the ratio of category 1 misuse was about 1% for all subject
groups, indicating that category 1 misuse was not a serious problem for the subjects in the
experiments. Hence in the rest of this article we focus on category 2 mistakes.
4.3 Appearance Ratios of Honorific Patterns
For each of the position labels and each subject group, we calculated the appearance
ratios of honorific patterns corresponding to the subjects answers and position labels
averaged over speech intentions. The most common and the next-most-common honorific
patterns in the answers of the younger subject groups (i.e., YM and YF) and older subject
groups (i.e., OM and OF) for each position label are shown in Table VI. In this table, the
letter "m" (male) or "f" (female) are attached to the honorific patterns when their
frequency ranking differs for male and female subjects. Values in parentheses attached to
the most frequently appearing honorific patterns show the appearance ratio (percentage)
for male and female subjects. For each column of rows whose position labels are SLAB,
SLBA, SALB, SABL, SBLA, SBAL, LSAB, LSBA, ASBL, and BSAL (where the most
frequently appearing honorific patterns vary for younger and older subjects), there are
significant differences in the ratios of honorific patterns for younger and older subjects
when the data for males and females is compounded in each age group. For example, the
appearance ratio for the honorific pattern 0000, corresponding to the position label SLAB
averaged over the subjects in each age group, differed significantly for younger and older
subjects.
We made the following discoveries about the most frequently appearing honorific
patterns in Table VI.
(1) The use of honorific patterns are usually the same for male and female subjects in
both the younger and older subject groups. The only exceptions are for position
labels LBAS and BALS for the younger subject groups; these position labels have
two alternative honorific patterns in Table III.
(2) The honorific patterns underlined in Table VI differ for the younger and older
subject groups; all these patterns are related to honorific titles. The other honorific
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
Using Japanese Honorific Expressions: A Psychological Study 157
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
158 T. Shirado et al.
patterns are the same for both the younger and the older subject groups, except for the
position labels LBAS, BALS, BLAS, and BASL, which show two alternative honorific
patterns in Table III.Comparing Table III with Table VI, we also noticed the following.
(1) For any of the position labels in Table III with only one honorific pattern (i.e.,
position labels other than LBAS, BALS, BLAS, and BASL), the corresponding
honorific pattern is the same as the most frequently appearing honorific pattern
that corresponds to the position label in Table VI for the younger subjects.
However, this is not necessarily true for the older subjects; where there are
differences related to honorific titles (i.e., related to the underlined honorific
patterns in the OM and OF column).
(2) For any of the position labels in Table III with two alternative honorific patterns
(i.e., LBAS, BALS, BLAS, or BASL), either of the alternative patterns from
Table III is the most or next-most-frequently appearing honorific pattern
corresponding to the position label in Table VI for both the younger and the older
groups.
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
Using Japanese Honorific Expressions: A Psychological Study 159
used by the female (OF) group, and its appearance ratio is almost the same as that of
1111 (there is only 1% difference between them); in addition, 1102 is the second most
frequently appearing honorific pattern used by all the older subjects.
Therefore, Table VII represents the knowledge shared in common by the OM and OF
subject groups.
Table VII (containing the knowledge of most of the older subjects) differs from Table
V (textbook knowledge) in the following ways.
(1) An honorific title does not follow A ("A" indicates a person's name); A=0 when the
social position of person A is lower than those of the speaker, the listener, and
person B (S>A L>A B>A). An honorific title follows A; A=1 when the social
position of person A is higher than those of the speaker, the listener, or person B
(A>S A>L A>B).
(2) An honorific title does not follow B ("B" indicates a person's name); B=0 when the
social position of person B is lower than those of the speaker, the listener, and
person B (S>B L>B A>B). An honorific title follows B; B=1 when the social
position of person B is higher than those of the speaker, the listener, or person A
(B>S B>L B>A).
(3) The object honorific expression is used secondarily; P=2 (secondary) is the
predicate when the social position of person B is higher than that of person A and
the social position of person A is higher than that of the speaker (B>A>S).
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
160 T. Shirado et al.
The hit ratio for each summarized consistency table for the honorific element P was
calculated in two ways.
The hit ratio[P=0], the hit ratio[P=1], and the hit ratio[P=2] were calculated by Eq. (1)',
and the subtotal hit ratio[P] was calculated by Eq. (2)', where Eqs. (1)' and (2)' are the
same as Eqs. (1) and (2), respectively, except that the answers for the partial relative
position B>A>S are ignored and v =0, 1, 2 in both equations.
The hit ratio[P=1 or w ], where w is 3 or 2 for Tables V and VII, respectively, was
calculated as
( all(PRP[P= v ])+all(B>A>S)),
v
Table VIII shows all the hit ratios calculated by the above equations. The first column
shows the summarized consistency table used for each calculation, which was common
for honorific elements P (except the partial relative position B>A>S) and E. The second
column shows the values of each honorific element, where A (Total), B (Total), E (Total),
and P (Total) are the total hit ratios for honorific elements A, B, E, and P, respectively,
and P (Subtotal) is the subtotal hit ratio for honorific element P. The third column shows
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
Using Japanese Honorific Expressions: A Psychological Study 161
the partial relative positions corresponding to the values of honorific elements in the
second column, defined by the summarized consistency table in the first column. The
fourth to seventh columns show the hit ratios for each subject group. The details of the hit
ratios for P=1, 2, or 3 for the partial relative position B>A>S are shown in parentheses.
The bottom two lines of Table VIII (where "Mean" is shown in the second column)
show the total hit ratio averaged over the honorific elements for Tables V and VII,
respectively.
4.6 Detailed Discussions of Table VIII
To allow a detailed discussion of Table VIII results, we also calculated the appearance
ratio of each value in the honorific patterns corresponding to the answers given by
subjects in each subject group (Table IX). Table IX thus represents the general tendency
of each subject group regardless of relative social position.
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
162 T. Shirado et al.
Table IX. Appearance Ratio for Each Value of Each Honorific Element
Value of Subject group
honorific element YM YF OM OF
A=0 0.444 0.406 0.390 0.353
A=1 0.556 0.594 0.610 0.647
B=0 0.440 0.402 0.386 0.350
B=1 0.560 0.598 0.614 0.650
E=0 0.520 0.488 0.507 0.445
E=1 0.480 0.512 0.493 0.555
P=0 0.400 0.407 0.424 0.400
P=1 0.325 0.354 0.357 0.362
P=2 0.225 0.203 0.186 0.207
P=3 0.050 0.036 0.033 0.031
The notable points of Table VIII follow: The hit ratios of A=0 and A=1 are almost the
same as those of B=0 and B=1, respectively, for any subject group in any summarized
consistency table. This suggests that the strategy chosen by subjects regarding the usage
of honorific titles for a person does not depend on whether the person is the subject or the
object of the predicate.
The hit ratios of the OF subject group for A=0 (0.388 in Table V and 0.673 in
Table VII); B=0 (0.390 in Table V and 0.673 in Table VII); and E= 0 (0.626) are
the lowest among the subject groups. This suggests that the experiments were
affected by a gender gap, in that the female subjects tended to use polite words
more than male subjects did [Brown et al. 1987]. In fact, Table IX shows that the
appearance ratios for A=1, B=1, E=1, and P=1 of the OF group were the highest (so
the ratios for A=0, B=0, and E=0 for the OF group were the lowest) among the
subject groups, indicting that the OF subject group in general tended to use polite
expressions. The reason the YF subject group did not show such a tendency is
possibly because the younger female subjects were university students who had too
little experience in society, and thus were not as affected by the gender gap.
The hit ratio of the OM subject group for P=2 for the partial relative position
B>S>A (0.554) was the lowest among the subject groups. To find the reason for
this, we calculated the appearance ratios of answers corresponding to P=0, P=1, and
P=3. Table X shows the results with the hit ratio (i.e., the appearance ratio of P=2).
The appearance ratio for P=0 (i.e., the predicate is neither respectful nor humble) in
the answers of the OM group for the partial relative position B>S>A was 0.425,
much higher than the ratio for the other subject groups (0.140, 0.136, and 0.162).
These results suggest that the OM group had a stronger tendency to believe that no
honorific expression should be used to express the action of a person (person A in
our experiments) whose social position was lower than that of the speaker.
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
Using Japanese Honorific Expressions: A Psychological Study 163
The hit ratios of the younger subjects for the partial relative position B>A>S are not high
in both consistency tables (0.531 (YM) and 0.529 (YF) in Table V and 0.537 (YM) and
0.651 (YF) in Table VII, due to the dispersion for the partial relative position, B>A>S, in
the answers of the younger subjects. We can see that the component hit ratios P=1, P=2,
and P=3 (shown in parentheses) of the younger subjects are 0.237, 0.300, and 0.294,
respectively, for the YM group, and 0.295, 0.356, and 0.234, respectively, for the YF
group. These features suggest that the younger people had not acquired a specific strategy
regarding the honorific type of predicate to use for the partial relative position, B>A>S.
The mean hit ratios in Tables V and VII range from 0.763 to 0.915 and 0.850 to 0.916,
respectively. The principal reason for misses is probably that the proposed model was
minimal; that is, it was not sufficiently complex to describe the knowledge of each
subject group in detail. For example, we would expect the mean hit ratios to rise if the
details of the strategy employed by the OF subject group for honorific elements A, B, and
E were analyzed and incorporated into the present model. On the other hand, excessive
specialization would lead to a lack of universality. Therefore, it is meaningful that 76% to
92% of the knowledge of the subjects regarding each honorific element can be
represented by a minimal model such as ours.
5. CONCLUSION
We investigated the knowledge regarding normative honorific expressions provided by
textbooks and that demonstrated by subjects during psychological experiments. The 40
subjects who participated in the experiments were divided into 4 groups according to age
(younger/older) and gender (male/female).
The experimental results reveal that knowledge regarding normative honorific
expressions in performance provided by textbooks is similar to that displayed by the
younger subjects, but differed from that shown by older subjects. The knowledge
displayed by older subjects was more complicated than that in textbooks or that shown by
younger subjects.
The model will be useful for developing computer-aided systems for teaching how
honorific expressions should be used. The rules defined in both Table V (the summarized
consistency table obtained from textbooks) and Table VII (the summarized consistency
table obtained from older subjects), concerning honorific elements P (except where the
partial social position is B>A>S) and E in both tables, can be used as normative rules for
constructing a system to point out misuse of honorific expressions. Where the partial
relative position is B>A>S, P= 1 is most recommended for use as the normative rule,
although P= 2 or 3 is admissible. Regarding honorific elements A and B, the rules in
Table V and Table VII are (alternatively) recommended, depending upon the educational
purpose. The rules defined in Table V may be suitable for learning basic or strict norms
of honorific expressions, whereas those defined in Table VII may be suitable for learning
higher-level expressions that are more useful in practical conversations.
The model still needs to be improved so that it can deal with situations where peoples
social positions in spoken sentences are almost equal, since such situations are likely to
occur frequently in practical conversations. Additional psychological experiments will be
required to achieve this
REFERENCES
BROWN, P. AND LEVINSON, S. 1987. Politeness - Some Universals of Language Usage. Cambridge University
Press, Cambridge, UK.
GOFFMAN, E. 1967. Interaction Ritual: Essays on Face to Face Behavior. Anchor Books, Garden City, NY.
GRUNDY, P. 1995. Doing Pragmatics. Oxford University Press, London.
HAGINO, N. 2001. Minasan-Korega-Keigo-desuyo. Riyon sha (in Japanese).
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.
164 T. Shirado et al.
HAYASHI, S. AND MINAMI, F. (EDS). 1974. Keigo-kouza 8 Sekaino-Keigo. Meiji Shoin (in Japanese).
HORIKAWA, N. AND HAYASHI S. 1969. Keigo-Yorei-Chushin Guide. Meiji Shoin (in Japanese).
HOSHINO, K. AND MARUYAMA, N. 1993. Nihongo-no-Hyogen. Keibun sha (in Japanese).
KARNAUGH, M. 1953. The map method for synthesis of combinational logic circuits. Trans. AIEE 72, 9, 593-
599.
KIKUCHI, Y. 1996. Keigo. Maruzen (in Japanese).
KIKUCHI, Y. 1997. Keigo Sai-Nyumon. Koudan-Sha (in Japanese).
MATSUMOTO, Y. 1999. http://chasen.aist-nara.ac.jp/hiki/ChaSen/). Japanese morphological analysis system
ChaSen version 2.3.3 Users Manual.
MATSUOKA, T. AND TAKUBO, Y. 1989. Kiso-Nihongo-Bunpo. Kuroshio Shuppan (in Japanese).
MIZUTANI, S. 1995. Taigu-Hyogen-Teiyo. Institute of Behavioral Sciences (in Japanese).
MORIYAMA, T. 2000. Kokokara-Hajimaru-Nihongo-Bunpo. Shobo Hitsuji (in Japanese).
SUZUKI, K. AND HAYASHI, D. 1984. Kenkyu-Shiryo Nihon-go Bunpo 9. Meiji Shoin (in Japanese).
WATANABE, M. 1971. Kokugo-Koubun-Ron. Hanawa-shobo (in Japanese).
ACM Transactions on Asian Language Information Processing, Vol. 5, No. 2, June 2006.