ArticlePDF Available

Abstract and Figures

This paper is an exploration in the semantics and pragmatics of linguistic feedback, i. e. linguistic mechanisms which enable the participants in spoken interaction to exchange information about basic communicative functions, such as contact, perception, understanding, and attitudinal reactions to the communicated content. Special attention is given to the type of reaction conveyed by feedback utterances, the communicative status of the information conveyed (i. e. the level of awareness and intentionality of the communicating sender), and the context sensitivity of feedback expressions. With regard to context sensitivty, which is one of the most characteristic features of feedback expressions, the discussion focuses on the way in which the type of speech act (mood), the factual polarity, and the information status of the preceding utterance influence the interpretation of feedback utterances. The different content dimensions are exemplified by data from recorded dialogues and by data given through linguistic intuition. Finally, two different ways of formalizing the analysis are examined, one using attribute-value matrices and one based on the theory of situation semantics.
Content may be subject to copyright.
1
Journal of Semantics, 1993, Vol. 9. No.1.
Also in Gothenburg Papers in Theoretical Linguistics 64, Dept of Linguistics,
Göteborg University, 1992.
On the Semantics and Pragmatics of Linguistic
Feedback
Jens Allwood, Joakim Nivre, and Elisabeth Ahlsén
University of Göteborg
Abstract
This paper is an exploration in the semantics and pragmatics of linguistic feedback,
i.e., linguistic mechanisms which enable the participants in spoken interaction to
exchange information about basic communicative functions, such as contact,
perception, understanding, and attitudinal reactions to the communicated content.
Special attention is given to the type of reaction conveyed by feedback utterances,
the communicative status of the information conveyed (i. e., the level of awareness
and intentionality of the communicating sender), and the context sensitivity of
feedback expressions. With regard to context sensitivity, which is one of the most
characteristic features of feedback expressions, the discussion focuses on the way in
which the type of speech act (mood), the factual polarity and the information status
of the preceding utterance influence the interpretation of feedback utterances. The
different content dimensions are exemplified by data from recorded dialogues and by
data given through linguistic intuition. Finally, two different ways of formalizing the
analysis are examined, one using attribute-value matrices and one based on the
theory of situation semantics.
___________________________________________________________________
Authors' address:
Jens Allwood, Joakim Nivre, and Elisabeth Ahlsén
Department of Lingustics
University of Göteborg
Box 200
S-405 30 Göteborg
Sweden
2
1. Purpose
The purpose of this paper is to present a sketch of a semantic/pragmatic account of
linguistic feedback mechanisms in spoken interaction. After an initial account and
exemplification of the relevant semantic/pragmatic features has been made, two
attempts at formalizing these are presented and discussed.
2. Background
2. 1. Analytic components of communication
Direct human face-to-face communication can be seen as the product of analytically
separable, interdependent functional subsystems.
In Allwood, Nivre and Ahlsén (1990), it was suggested that, at least for some
purposes, the following three overriding functions might be fruitful to consider for
speech production and speech perception.
(i) Speech management functions, i.e., the linguistic processes and mechanisms
whereby a speaker manages his or her own linguistic contributions to a
communicative interaction, involving phenomena that have sometimes been
described as “planning”, “editing”, “(self)repair” etc.
(ii) Interactive functions, i.e., linguistic processes and mechanisms whereby the
speakers manage the flow of interaction. (Feedback mechanisms, the topic of
this paper, is an example of an interactive subsystem.)
(iii) Focussed or main message functions, i.e., linguistic processes and
mechanisms whereby speakers manage to communicate information which is
not immediately connected with management of their own speech or the
interaction at hand. Focussed or main message functions, thus, include most
of what is commonly described in grammatical theory and can be
operationally defined as that which is contained in an utterance when those
parts that are devoted to speech management or interactive functions have
been substracted.
Speech management, interactive functions and focussed/main message functions can
further be analytically subdivided into subsystems and subsystems of subsystems,
3
characterized by different functions. Interaction functions can, for example, be
subdivided into mechanisms for:
(i) sequencing (of activities and subactivities, communicative acts and/or topics)
(ii) turntaking
(iii) giving and eliciting feedback.
The literature on conversation analysis and discourse analysis (see e.g. Levinson
1985 or Brown and Yule 1983) contains much discussion of the former two types of
mechanisms, whereas there has been less discussion of feedback (cf. Allwood 1988a,
1988b). This paper is intended as a contribution to the further exploration of
linguistic feedback mechanisms, especially with regard to the semantic/pragmatic
functions of such mechanisms.1
2.2. Linguistic feedback: basic functions
The term feedback originates in cybernetics (Wiener 1948), where it is used to
denote processes by which a control unit gets information about the effects and
consequences of its actions.
Here we are concerned with linguistic (interindividual) feedback (Allwood 1979,
1988a, 1988b, 1988c), i.e., linguistic mechanisms which enable the participants of a
conversation to exchange information about four basic communicative functions,
which are essential in human direct face-to-face communication. These functions
are:
(i) contact (i.e., whether the interlocutor is willing and able to continue the
interaction)
(ii) perception (i.e., whether the interlocutor is willing and able to perceive the
message)
(iii) understanding (i.e., whether the interlocutor is willing and able to understand
the message)
1 As can he seen. we are here making no attempt to distinguish semantics from pragmatics. This is
so because we believe that such a distinction runs into serious practical and theoretical difficulties
(cf. Allwood 1981).
4
(iv) attitudinal reactions (i.e. whether the interlocutor is willing and able to react
and (adequately) respond to the message, specifically whether he/she accepts
or rejects it).
These four basic functions of linguistic feedback arise from four basic requirements
of human communication. First, communication requires that at least two agents are
willing and able to communicate. Second, communication requires that the receiving
agent is willing and able to perceive the behavioral or other means whereby the
sending agent is displaying or signalling information. Third, communication requires
that the receiving agent is willing and able to understand the content that the sender
is displaying or signalling. It is also often helpful if the receiver can perceive and
understand various types of indicated information.2 Finally, communication requires
that the receiving agent is willing and able to react attitudinally and behaviorally to
various aspects of the content that the sender is displaying or signalling. Again, it is
sometimes beneficial for communication, if the receiver also reacts to indicated
information. Certain conventional features of the displayed or signalled content here
seem particularly important for the interpretation of the content of feedback
expressions. Among these are polarity (positive or negative) and mood
(conventionally signalled evocative intention; cf. Allwood 1978).
Every language appears to have conventionalized means (verbal and prosodic means
as well as body movements) for giving and eliciting information about the four basic
communicative functions. Linguistic feedback mechanisms on a primary level
usually involve very short morphemes (yes, no, m), or basic mechanisms such as
repetition, simple body movements (head nods, head shakes) in combination, on a
secondary level, with fairly simple phonological, morphological and syntactic
operations for modifying and expanding the primary feedback expressions.
Earlier studies that have discussed feedback and related phenomena include Allwood
(1976, 1979, 1988a, 1988b), Anward (1986), Clark & Schaefer (1989), Ehlich
(1986), Fries (1952), Hellberg (1985), Heritage (1984), James (1972), Nivre (1991),
Schegloff (1982), Severinson-Eklundh (1986), Sigurd (1984), Yngve (1970).
Allwood (1988b) gives a taxonomy for the structure of linguistic feedback and, in
particular, describes the Swedish system. In the present paper, we want to focus on
the content features of linguistic feedback.
2 For the distinction between indicated, displayed and signalled information, see section 3.3 below.
5
3. Content features of feedback
3.1. Introduction
Although simple feedback words, like yes, no and m, are among the most frequent in
spoken language, a proper analysis of their semantic/pragmatic content seems to be
fairly complex and involve several different dimensions. One striking feature is, for
example, that these words involve a high degree of context dependence.
Below, we will first discuss four of these dimensions and exemplify them by data
from recorded dialogues and by data given through linguistic intuition. The examples
from recorded dialogues are all in Swedish (with English translations). In addition to
this, Swedish is used to exemplify distinctions which can not be found in English.
The four dimensions we will discuss are:
(i) Type of reaction to preceding communicative act
(ii) Communicative status
(iii) Context sensitivity to preceding communicative act, with regard to:
A. Type of speech act (mood)
B. Factual polarity
C. Information status
(iv) Evocative function.
3.2. Type of reaction to the preceding communicative act
The raison d'être of linguistic feedback mechanisms is the need to elicit and give
information about the basic communicative functions, i.e., continued contact,
perception, understanding and emotional/attitudinal reaction, in a sufficiently
unobtrusive way to allow communication to serve as an instrument for pursuing
various human activities. The linguistic feedback system is, in this way, an essential
instrument for successful communication of any type. Especially, it is an essential
instrument for the incrementality of communication, i.e., the step by step build up of
consensual joint understanding and attitudes. Feedback mechanisms are, thus, a
means for communication which in its turn is a means for pursuing a variety of other
human activities.
In our analysis of the content of feedback we are assuming that what we have called
the basic functions also define four basic dimensions in the reactions that
interlocutors have to each other's contributions in conversation. Feedback utterances,
thus, give information about one or several of the following types of reaction:
6
(i) contact - willingness and ability to continue interaction
(ii) perception - willingness and ability to perceive expression
and message
(iii) understanding - willingness and ability to understand expression
and message
(iv) (other) attitudinal reactions - willingness and ability to give other attitudinal
reactions to expression, message,or interlocutor.
Category (iv) has the word other in brackets, since contact, perception, and
understanding also involve attitudes, albeit of a very fundamental cognitive and
volitional sort. Category (iv), which we will mostly just refer to as attitudinal
reactions without other, is supposed to cover other attitudes such as acceptance,
non-acceptance, belief, disbelief, surprise, boredom, disappointment, enthusiasm,
etc. When it comes to the words yes and no and their synonyms, we believe that the
attitudes acceptance and nonacceptance are in focus and form a basis which can be
modified by added attitudinal reactions. We can, thus, accept with regret or with
enthusiasm by uttering the word yes with different types of prosody. In general, we
can say that feedback words differ from each other mainly with regard to what
attitude they signal, e.g., yes - acceptance, no - non-acceptance, great - appreciation/
enthusiasm, etc.
In example (1) below, ja (yes) has the functions of conveying continued contact,
perception and understanding as well as the attitudinal reaction acceptance.
(1) A: men efter tre år va de ju3 en härlig mylla
(but after three years you-know it was a lovely mould)
B: ja
(yes)
We can compare this to example (2), where B's weaker feedback utterance mm has
the same content with respect to contact, perception and understanding, but does not
necessarily convey the attitudinal reaction of acceptance of the veridicality of A's
statement.
3 The Swedish word ju appears in some of the examples in this paper. Ju has no exact translation
into English. It has the function of making what is stated appear as mutually known information.
This might depending on context variously he rendered as "you know" or "we know". For reasons
of idiomaticity. we have chosen to use the hyphenated expression you-know. in our translations
although this is not always a good equivalent. It should also he observed that ju is less salient and
weaker than you-know.
7
(2) A: ...ja kan få såna / / kontakter …kontakter mä universum jaa
(yes I can get such / / contacts … contacts with the universe yes)
B: MM
(mm)
One might, however, claim that mm still signals acceptance in the weaker sense of
accepting to continue, accepting the information in the preceding utterance as
perceived and understood and possibly also of accepting to take a stand on this
information.
3.3. Commununicative status
Like any other information communicated, feedback information concerning the
basic communicative functions can be given on many levels of awareness and
intentionality. This is so, whether the information is communicated by verbal or
bodily means. Although levels of awareness and intentionality almost certainly are a
matter of degree, we, in order to simplify matters somewhat, here distinguish three
levels from the point of view of the communicating sender (cf. Allwood 1976):
(i) Indicated information is information that the sender is not aware of, or
intending to convey. This information is mostly communicated by virtue of
the receiver's seeing it as an indexical (i.e., causal) sign.
(ii) Displayed information is information that the sender is intending to “show”
the receiver. The receiver does not, however, have to recognize this intention.
Display of information can be achieved through any of the three main
semiotic types of signs (indices, icons and symbols, cf. Peirce 1955).
(iii) Signalled information is information that the sender is “showing” the receiver
that he is displaying and, thus, intends the receiver to recognize as displayed.
Signalling can also be achieved through any of the three main semiotic types
of signs. In particular, however, we will regard ordinary linguistic expressions
(verbal symbols) as being signals by convention. Thus, a linguistic expression
like It's raining, when used conventionally, is intended to evoke the receiver's
recognition not merely that “it's raining” but that he/she is “being shown that
it's raining”.
The fact that linguistic expressions by convention are taken to be signals, does not,
however, imply that they are always actually used as signals. A symbol can also be
used to indicate and/or display its conventionally signalled content or some other
content. Compare the example discussed by Searle (1969), where an American
8
soldier, captured by the Italians in World War II, by quoting “kennst Du das Land,
wo die Zitronen blühen” (do you know the land where the lemons bloom) intends to
display to the Italians that he is German.
In order to illustrate the application of the dimension of communicative status to
linguistic feedback utterances we will consider the communicative status of some
examples from the recorded dialogues.
In examples (1) and (3) below, the communicative status of the feedback utterances
produced by B is not quite the same. In both cases the preceding utterance (produced
by A) is a declarative statement with positive polarity (i.e., it is not negated) and in
both cases the feedback utterance signals the acceptance function by use of ja (yes),
while it indicates continued contact as well as perception and understanding of the
preceding utterance. In example (1), however, the simple ja (yes) can merely be said
to indicate the attitude of belief, while in example (3), the more elaborated feedback
utterance rather signals belief, expressed through the indicative mood of the sentence
de e de ju (it is you know).
(1) A: men efter tre år va de ju en härlig mylla
(but after three years you-know it was a lovely mould)
B: ja
(yes)
(3) A: de e ju väldit faalit me kärnkraft
(it is very dangerous you-know with nuclear power)
B: ja // de e de ju
(yes // it is you-know)
3.4. Context sensitivity with regard to the preceding communicative act
3.4.1.Introduction
One way of analyzing the meaning of linguistic feedback expressions and
mechanisms is to say that they are characterized by a very abstract conventional type
content in combination with high degree of context sensitivity. For example, the
conventional type content of the three expressions yes, no, m, and ok can perhaps be
characterized in the following way:
yes - acceptance
no - rejection
n- confirmation
ok - agreement
9
The conventional occurrence content of the three expressions is, however, always
also a function of prosody and context. The function of prosody is mainly to
modulate attitudinal information. In some cases (cf. example 5 below), this can
affect the presupposed truth of the preceding utterance. Prosody will, however, not
be treated in any detail in this paper. As for context, table 1 below demonstrates the
influence of mood and polarity of the preceding utterance.
Table 1. Functions of yes, no, m, and ok in relation to the mood and polarity of
the preceding utterance.
Preceding
utterance Listener's
response
yes no m ok
Pos statement
It's raining Acceptance of
statement (Indicated
belief)
Rejection of
statement Confirmation of
understanding
(Indicated acceptance
of statement)
Agreement
(Acceptance of
what has been said
as a point of
departure, more or
less stipulatively)
Neg statement:
It isn't raining Ambiguous between
rejection of statement
(yes it is) and
acceptance of
statement (yes you
are right)
Acceptance of
statement
(Indicated belief)
Confirmation of
understanding Agreement
(Acceptance of
what has been aid
as a point of
departure …)
Pos yes-no
question:
Is it raining?
Commitment to
positive fact Commitment to
negative fact Confirmation of
understanding
(Indicated
commitment to
positive fact)
Agreement
(Acceptance of
implicit suggestion)
Neg yes-no
question:
Isn't it raining?
Commitment to
positive fact Commitment to
negative fact Confirmation of
understanding
(Indicated
commitment to
positive fact))
Agreement
(Acceptance of
implicit suggestion)
Pos request:
Open the door! Acceptance of
request Refusal of request Confirmation of
understanding
(Indicated acceptance
of request)
Agreement
Neg request:
Don't open the
door!
Unclear Acceptance of
request Confirmation of
understanding
(Indicated acceptance
of request)
Agreement
Pos offer:
Would you like
some tea?
Acceptance of offer Rejection of
offer Confirmation of
understanding
(Indicated acceptance
of offer)
Agreement
(Indicated
acceptance on the
grounds of what has
been said)
Neg offer
Wouldn't you
like some tea?
Acceptance of offer Rejection of
offer Confirmation of
understanding
(Indicated acceptance
of offer)
Agreement
(Indicated
acceptance on the
grounds of what has
been said)
10
We can see how context can change the occurrence content both with regard to
attitudinal reaction (from acceptance to non-acceptance) and with regard to
attitudinal object (e. g., from statement to offer). The table is somewhat unnatural in
that simple feedback expressions without pronominal indications of the objects of
acceptance and nonacceptance (yes it is, no it isn't) have been used. In the case of
yes, this leads to ambiguity after a negative statement (ambiguous between
acceptance of a negative statement and acceptance of the positive counterpart of the
negative statement = rejection) and unclarity after a negative request (yes I will(?),
yes I won't(?)).
We seem to have a sort of semantic field constituted by terms like yes, no, m, and ok
supported by attitudinal dimensions of meaning like agreement, confirmation,
acceptance and commitment. Each term is primarily focussed towards one or several
of these dimensions, but can, depending on context, simultaneously indicate or
display other compatible dimensions or even, with a change of focus, signal other
dimensions.
The latter might happen, for example, in language acquisition, when a language
learner who is yet not very proficient in the language he/she is learning uses the
vagueness of the notion of acceptance connected with the word yes in order to signal
acceptance of continued communication rather than acceptance of perceived and
understood content. What is really being signalled (displayed or indicated, as the
case may be) is willingness or agreement to continue communication rather than the
more stereotypical fullbodied notion of accepting the evocative intention of the
preceding utterance (communicative act). If the receiver of the yes is not fully
informed about the learner's nonproficiency, there is a clear risk that what the learner
is signalling (displaying, indicating) will be misunderstood.
Just like deictic expressions (I, you, here, there, now, then, etc.), feedback
expressions are, thus, highly dependent on context for a precise determination of
their meaning. However, just as is the case with deictic terms, this dependence is not
random, but in fact triggered by specific contextual parameters. As can be seen from
the discussion and examples above and further from the examples to be discussed
below, among the most important of these parameters are various features of the
immediately preceding communicative act:
(i) Type of speech act (mood)
(ii) Factual polarity
(hi) Information status.
11
3.4.2. Type of speech act (mood)
Table 2, which is extracted from table 1, illustrates the status of yes in different
contexts. More precisely, we can see that the object of acceptance is determined by
mood and speech act status. In the examples, we are making the assumption that
mood and speech act status are in harmony. When mood and speech act status
differ, increased degrees of freedom as to object of acceptance are introduced and
context seems to determine which is chosen.
Table 2. Effects of speech act status (mood) on feedback.
Preceding utterance Listener's reply Function
It's raining yes Acceptance of statement
It's raining yes Commitment to positive fact
Open the door! yes Acceptance of request
Would you like some
coffee? yes Acceptance of offer
We also see that the speech act status of the preceding communicative act can trigger
a change in the attitude signalled. A yes-no question can, at least in some cases, be
analyzed as a request for a commitment on the part of an interlocutor as to the
veridicality of some statement. A reply using yes or no will therefore indicate a
positive or negative commitment to an indicated fact and not merely acceptance of
this fact.
If we contrast examples (1) and (4), we see the partially different functions of ja/jaa
(yes) after a statement (example 1), where it conveys acceptance of the statement and
after a question (example 4), where it conveys commitment to a positive fact.
(1) A: men efter tre år va de ju en härlig mylla
(but after three years you-know it was a lovely mould)
B: ja
(yes)
(4) A: e ni klara då
(are you finished then)
B: jaa
(yes)
The vowel reduplication in jaa is one of the means whereby a speaker can show
increased commitment.
12
Further, if we take the meaning of yes and no to be acceptance and non-acceptance
(rejection), it might be tempting to assume that they, when following a statement,
like in the case above, always directly indicate acceptance or non-acceptance of this
statement. This is, however, an oversimplification as is shown by the example below.
(5) A: it's raining
B: oh no
Here, oh no, if pronounced in a short, matter of fact way, can indicate denial of the
statement. But consider instead the possibilities of pronouncing oh no with a
disappointed or surprised intonation. In such cases, B would presuppose the truth of
A's statement in order to signal his emotional non-acceptance of something he, all
the same, believes to be true.
The object of acceptance or non-acceptance contextually signalled by yes and no,
thus, does not merely depend on the status of the preceding communicative act but
also on what type of attitudinal reaction the feedback utterance signals. Attitudes
such as disappointment or surprise are factive and presuppose some fact towards
which they are directed. This presupposition seems to be upheld in the case above
and the nonacceptance instead to be used as an underpinning of the unpleasantness
or unexpectedness signalled by the word oh in conjunction with the prosodic
expression of disappointment or surprise.
3.4.3. Factual polarity
If we look at examples (6) and (7) below, we can see how the factual polarity of the
preceding communicative act affects the function of the feedback utterance Consider
the use of nä/nej (no) in examples (6) and (7) below.
(6) A: de kan ju inte va för fiskarnas skull va
(it couldn't be for the sake of the fish you-know)
B: nä
(no)
In example (6) the preceding statement has negative polarity and the function of the
negative feedback utterance is acceptance. In example (7), on the other hand, the
preceding statement has positive polarity and the function of the negative feedback is
nonacceptance.
(7) A: så går naturen under me tekniken
(like that nature perishes with technology)
13
B: NEJ / / de växer upp annat då vet du
(no / / other things grow up you know)
Table 3, which is also extracted from table 1, illustrates the role of the factual
polarity of the preceding utterance. As we can see, the polarity of the preceding
utterance affects the attitude expressed by a yes or a no. If a statement preceding a
yes is positive, the yes signals acceptance of the statement. If the statement, however,
is negative, the yes can signal objection and rejection of the proposed negative
statement. Normally, however, this function has to be supported by the positive
pronominal reformulation it is. Likewise a no following a positive statement signals
rejection of the statement, but following a negative statement it signals acceptance.
The polarity of the preceding utterance, thus, seems to have a particularly drastic
effect on the attitude signalled by a yes or a no.
Table 3. Effects of the factual polarity of the preceding utterance on feedback.
Preceding
utterance Listener's
response
yes (it is) no (it isn't)
Pos statement
It's raining Acceptance of
statement (Indicated
belief)
Rejection of statement
Neg statement:
It isn't raining Rejection of
statement Acceptance of
statement (Indicated
belief)
Pos yes-no question:
Is it raining? Commitment to
positive fact Commitment to
negative fact
Neg yes-no question:
Isn't it raining? Commitment to
positive fact Commitment to
negative fact
yes (I will) no (I won't)
Pos request:
Open the door! Acceptance of
request Refusal of request
Neg request:
Don't open the door! Rejection of request
(Defiance) Acceptance of request
yes (I would) no (I wouldn't)
Pos offer:
Would you like some
tea?
Acceptance of offer Rejection of offer
(Declination)
Neg offer
Wouldn't you like
some tea?
Acceptance of offer Rejection of offer
(Declination)
14
If we look somewhat more closely at table 3, we see that statements and requests
seem to pattern one way and yes-no questions and offers a slightly different way
with regard to the effect of their polarity on the content of yes and no. In the case of
statements and requests, positive polarity results in acceptance (yes) and rejection
(no), while negative polarity results in the converse rejection (yes) and acceptance
(no). What seems to be accepted or rejected in the case of requests is the task of
carrying out the request, while following statements, acceptance (yes and no)
ambiguously can concern what might be termed provisional acceptance or it might
concern a more fullbodied acceptance and integration into one's own system of
beliefs. Rejection following statements seems in the case of both yes and no to signal
commitment to fact with a polarity opposite the one indicated by the statement.
In the case of preceding yes-no questions and offers (which in the examples given
here also have the form of yes-no questions), change of polarity does not seem to
have the same effect, so that yes signals commitment to positive fact and no signals
commitment to negative fact, regardless of the polarity of the preceding utterance.
In order to maintain the same analysis for all four contexts we could say that the yes,
where it follows a negative yes-no question or offer (since negation to be relevant
seems to presuppose a positive expected state of affairs which is denied) signals
acceptance of this expected positive state of affairs. A no would signal rejection of
this expected positive state of affairs.
Another alternative to maintain the same analysis for all four contexts would be to
claim that yes always involves commitment to positive fact and no commitment to
negative fact. This analysis would, in fact, also work for yes and no following
statements and requests, where, for example, a yes signalling commitment to positive
fact following a negative statement would indicate objection or rejection of the claim
made and when following a positive statement would indicate acceptance or
agreement. Even if the analysis of yes and no as signalling commitment to positive
and negative facts, respectively, perhaps seems somewhat simpler than the
acceptance/non-acceptance analysis, it runs into problems with the case discussed in
example (5), i. e. where no is preceded by oh and pronounced with an intonation
conveying disappointment. Such a response seems to presuppose the correctness of
the speaker's claim, but signal the listener's emotive, conative non-acceptance.
Whichever analysis is chosen, it is, however, clear, that the attitude expressed by a
yes or a no requires consideration of the polarity of the immediately preceding
utterance in order to be determined.
15
In some languages, such as Swedish and German, the analysis just proposed for yes
and no in English would have to be made somewhat more complicated in order to
accommodate the fact that these languages have a special morpheme jo (Swedish)
and doch (German) which is used instead of yes in all the cases following an
utterance with negative polarity. So for Swedish and German one could therefore
suggest that the meaning of ja (the same word in both languages) is to accept to carry
out what the evocative function of a preceding positive utterance signals. In the case
of statements, yes-no questions, and yes-no offers, the ja furthermore often “delivers
the goods”, i.e., provides a commitment to one of the indicated alternatives. In the
case of requests, this is usually not possible since mostly nonverbal action going
beyond a simple yes is required to “deliver the goods”.
The function of jo and doch would, when following an utterance with negative
polarity, instead be to assert commitment to a positive corresponding state of affairs
opposite to that indicated by the preceding utterance. The Swedish and German
distinction between ja - jo and ja - doch would thus separate acceptance of a positive
state of affairs from commitment to a positive state of affairs as a reaction to an
utterance where this state of affairs has been given negative polarity. In English, yes
is instead polysemic with regard to these functions. Other languages, such as
Russian, offer a further modification of the analysis. The acceptance function of da
(yes) has been extended so that not only positive facts can be accepted, but also
negative facts. Consider the following example.
(8) A: nie idjot dozhd
B: da
B's utterance in the Russian example (8) signals acceptance of the fact that it is not
raining. Negative questions, requests and offers seem to function similarly, so that da
can be used to signal acceptance of a negative state of affairs. In English, the word
mm can be used in a similar way, the difference being that mm indicates rather than
signals acceptance.
3.4.4. Information status
A third feature of an utterance preceding a yes or a no that seems important both for
the actual morphological and phonological realization of yes or no and for their
interpretation is the information status that the utterance has for the listener, i.e., for
the person giving the feedback. Compare examples (9), (10), and (11).
(9) A: det regnar
(it's raining)
16
B: ja det gör det ja
(yes it does it yes)
(10) A: det regnar inte
(it's not raining)
B: nä det gör det inte nä
(no it does it not no)
A: det regnar
(it's raining)
B: *nä det gör det inte nä
(no it does it not no)
In example (9), the “sandwich” positioning of the ja before and after the pronominal
reassertion of the preceding statement serves to signal that the listener has been
reminded of something he/she already knew. The corresponding “sandwich”
construction with no can therefore be used after a negated statement, as in example
(10), only when it signals that B is reminded of a negative fact that he accepts as
true. It cannot be used in order to object to a positive statement, as in example (11).
If, in example (9), B had responded by ja ja, which could be regarded as an
abbreviated version of ja det gör det ja, the signalled meaning would have been
something like yes, I know, without the indication of having been reminded. If B had
responded by jaså (oh (really)), this would instead have signalled that the fact
mentioned by A was new to B, thus not something he was reminded of or already
knew. In fact, this feature of jaså (oh) can be ironically exploited in Swedish by
speakers who say jaså in order to indicate to their interlocutor that what they are
hearing is perhaps not so new and interesting as their interlocutor would like to
imagine.
Another operation on information status can be achieved by the use of jaha (oh)
which in example (9) could have been used to signal that B accepts that A says det
regnar as a fact, which is ambiguous between taking A's uttering something as a fact
and taking the state of affairs indicated by A as a fact. This ambiguity is brought out
in examples such as jaha, det är vad du säger (oh, that's what you say), jaha det är
vad du tror (oh, that's what you think) or jaha, då får vi ta med oss paraply (oh, then
we have to take an umbrella).
As we have seen, there are various means for making a feedback utterance indicate,
display or signal the information status of the preceding utterance in relation to the
17
person who gives feedback. In example (12) below, the use of the negative
morpheme (no), as a reaction to a preceding positive statement, as well as the
lengthening of the morpheme (no) by the added vowel -e, makes the utterance
display an attitude of surprise and thereby indicate that the information status of the
preceding utterance is new rather than given or known. In particular, as already
discussed in section 3.4.2., B is not denying the veridicality of A's statement.
(12) A: så ja har tomatlådor där å ja brukar få ett par hundra tomater
(so 1 have tomato boxes there and I usually get hundreds of
tomatoes)
B: näe
(no)
Another example, where the information of the preceding communicative act is
perceived as new, by virtue of the feedback utterance is example (13) below.
3.5. Evocative function
Feedback utterances conveying that the listener (B in our examples) is surprised and
that the information in the preceding utterance is new to him/her, often also have an
evocative function, i.e., they place an obligation on the current speaker (A) to react,
in his turn, and give feedback to B's feedback. Thus, B's jasså, in example (13),
displays surprise which leads A to reaffirm.
(13) A: å karamellpapprena dom kommer i i // i den där papperskorgen sen
(and the candy papers they get into into // into that waste paper
basket then)
B: jasså
(really)
A: jaa
(yes)
B: de va ovanlit
(that's unusual)
The word jasså (really) displays surprise and indicates that the preceding utterance
contains new information. An additional rising intonation can make this function
even stronger. As we can see, A responds with a feedback utterance jaa (yes), where
the added -a gives the utterance emphasis, i.e., A reaffirms his own preceding
18
utterance. B then continues de va ovanlit (that's unusual), which displays her
continued attitude of surprise.
In a somewhat wider sense of evocative, of course, every utterance containing only a
single feedback word could be said to evoke the continuation of the conversation.
Consideration of the evocative function of feedback, thus, connects it to the basic
function we have above referred to as ability and willingness to continue a
communicative interaction. By uttering a feedback word a speaker simultaneously
indicates willingness and ability to continue and willingness and ability to let the
other speaker continue.
4. Formalizing content features of feedback
4.1. Introduction
In this section, we want to explore the possibility of formalizing the analysis of
content features presented in section 3. In doing this we will develop two different
kinds of formalization, one using attribute-value matrices and the other based on the
theory of situation semantics.
4.1.1. Attribute-value structures
The first kind of formalization simply consists in using attribute-value matrices to
represent bundles of content features associated with linguistic expressions. Besides
offering a compact and yet perspicuous notation, the use of attribute-value matrices
(or feature structures, as they are sometimes called) potentially gives us a
unification-based formalism,4 which may be useful if you want to describe how the
occurrence content of a particular feedback utterance is constructed by combining a
type content with features of the context. (This is a problem that we will not really
pursue in this paper, however.)
4.1.2. Situation semantics
The second attempt at formalization is couched in the framework of situation
semantics (Barwise & Perry 1983, 1985; Barwise 1989). Within that theory, the
occurrence content P of a linguistic utterance is regarded as a function of two
parameters: the expression (type) S which is used, and the embedding circumstances
4 For an introduction to unification-based formalisms and their use in syntax and semantics. see
Shieber 1986, Pollard & Sag 1987, Johnson 1988.
19
(or context) c in which S occurs. This is expressed in the following semantical
“equation” (cf. Barwise 1986b):
(14) C(S, c) = P
The occurrence content of an expression S in a context c (as well as the content of
any other information-carrying event) is generally taken to be the information that
there exists a (real) situation of a certain type (i. e. that a certain type of situation is
realized). The content is therefore modelled in situation semantics with a situation
type (cf. Barwise 1986a, Israel & Perry 1988). A situation type T is defined by a
(possibly parametric) infon (or state of affairs) i, which is called the conditioning
infon of T. A particular situation s is of type T if and only if it supports i (or an
instance of i, if i is parametric), i.e. if and only if i is a fact in s. A situation-type T
with conditioning infon i is represented as in (15), where s is a situation variable and
the whole expression is read as “the type of situation s where i holds”.
(15) [ s | i ]
For example, if an utterance of the sentence It is raining in a certain context c has the
content “it is raining at a certain spatiotemporal location l" (say, the utterance
location), then we may express this as follows:
(16) C(It is raining, c) = [ s ] <<at l; raining; l>> ]
The picture sketched so far is oversimplified in (at least) one important respect. In
reality, the content of a linguistic utterance is not a function of two but of three
parameters. In addition to the expression used and the embedding circumstances, we
have to consider the set of constraints (law-like regularities such as linguistic
conventions) with respect to which the utterance interpreted, as the examples in (17)
make clear:
(17) a. C(Swedish, /næ/, c) = “no”
b . C(Greek, /næ/, c) = “yes”
Since we will only be concerned with one language (Swedish) in the formalized
examples below, we will generally suppress the constraint parameter in the
representations and continue to represent the content of linguistic utterances as a
function of only two parameters: expression and content. It is important to keep in
mind, however, that the content we attribute to particular utterances is always
dependent on a particular set of constraints (especially linguistic conventions).
20
4.2. Type of reaction to the preceding communicative act
Information about the basic communicative functions (contact, perception,
understanding, attitudinal reactions) can be represented in attribute-value format
using the four attributes CONTACT, PERCEPTION, UNDERSTANDING and ATTITUDE,
where the attribute ATTITUDE, as noted above, can be read as short for OTHER
ATTITUDE, since CONTACT, PERCEPTION and UNDERSTANDING also involve attitudes
(cf. section 3.1).
The first three will be treated here as binary features (with possible values + and -),
although this is an oversimplification. The ATTITUDE feature, by contrast, takes as its
value a complex feature structure containing information about the attitudinal
reactions which are present in different cases. This complex feature structure thus
contains a selection from a set of binary features ACCEPT, REJECT, BELIEF,
AGREEMENT, SURPRISE, etc.5
In sum, then, we need at least the following collection of features to represent type of
reaction to preceding communicative act:
(18) Feature Type of value
CONTACT BOOLEAN
PERCEPTION BOOLEAN
UNDERSTANDING BOOLEAN
ATTITUDE COMPLEX
ACCEPT BOOLEAN
REJECT BOOLEAN
BELIEF BOOLEAN
AGREEMENT BOOLEAN
SURPRISE BOOLEAN
The content of the feedback utterances ja and mm in examples (1) and (2) (repeated
below for convenience) can now be (partially) represented as in (19) and (20),
respectively.
(1) A: men efter tre år va de ju en härlig mylla
(but after three years you-know it was a lovely mould)
5 Here it is even more of an oversimplification to use simple binary features, and for two reasons.
First, the object of the attitudes may be different from one case to another, although it will be
assumed here that the object is always some feature of the content of the preceding utterance.
Second, the strength of the attitudes may vary, surprise, for example, may be expressed in different
degrees, although it will he treated here as a simple yes-no matter.
21
B: ja
(yes)
(19) [CONTACT +1
[PERCEPTION +]
[UNDERSTANDING +]
[ATTITUDE
[ACCEPT +] ]
(2) A: ... ja kan få såna / / ääh kontakter....kontakter mä universum jaa
( ... yes I can get such / / eeh contacts ... contacts with the
universe yes)
B: mm
(mm)
(20) [CONTACT +]
[PERCEPTION +]
[UNDERSTANDING +]
If we turn to situation semantics, we must remember first that the content of a
feedback utterance will be represented as a situation type (cf. section 4.1.2.). In most
cases, the conditioning infon of this type will be a complex one, consisting of a
conjunction of atomic infons (we will use the notation '& i1, ... ii' to denote the
conjunction of the infons i1, ... ii). For example, the content of the feedback
utterances in example (1) and (2) can be represented as in (21) and (22):
(21) C(ja, c1) = [s | & <<at l1; willing-to-continue, B1; 1>>,
<<at l1; perceive, B1, u1; 1 >>,
<<at l1; understand, B1, u1, P1; 1>>,
<<at l1; accept, B1, P1; 1>> ]
(22) C(mm, c2) = [s | & <<at l2; willing-to-continue, B2; 1 >>,
<<at l2; perceive, B2, u2; 1 >>,
<<at l2; understand, B2, u2, P2; 1 >> ]
We see that the conditioning infons in both cases are conjunctions of (possible) facts
about the speaker Bi and his willingness to continue, his perception of the preceding
utterance ui, his understanding of the content Pi of the preceding utterance ui, and (in
21 but not in 22) his acceptance of the communicated content Pi.
22
So far, we have not made any attempt to capture the influence of context in the
interpretation of feedback utterances. For example, in (21) there is no indication of
how the location l1, individual B1 utterance ul and content P1, (which are constituents
of the conditioning infon) are picked out from the context (and the context itself is
only represented by the symbol cl). We will return to this problem in section 4.4.
below.
4.3. Communicative status
Communicative status can be introduced into our attribute-value notation by means
of three complex-valued attributes INDICATE, DISPLAY and SIGNAL, which take
as their values feature structures representing the information which is indicated,
displayed or signalled, respectively. Their use is illustrated in (23), which is a richer
representation of the content of ja in example (1) than the one given in the preceding
section, and (24), which represents the content of ja de e de ju in example (3),
repeated below for convenience.
(23) [INDICATE [CONTACT +]
[PERCEPTION +]
[UNDERSTANDING +]
[ATTITUDE [BELIEF +] ] ]
[SIGNAL
[ATTITUDE
[ACCEPT +] ] ]
(3) A: de e ju väldit faalit me kärnkraft
(it is very dangerous you-know with nuclear power)
B: ja / / de e de ju
(yes / / it is you-know)
(24) [INDICATE
[CONTACT +]
[PERCEPTION +1
[UNDERSTANDING +]
[SIGNAL
[ATTITUDE
[ACCEPT +]
[BELIEF +] ] ]
23
In situation semantics, the notion of communicative status can be captured in
different ways. Here we approach the problem simply by dividing the content of an
utterance into three parts, namely indicated content, displayed content and signalled
content, which we will represent as C1(S, c), CD(S, c), and Cs(S, c), respectively.
Thus, we assume that the following equation holds (for arbitrary expressionsS and
contexts c):
(25) C(S, c = C1 (S, c + CD(S, c + CS(S, c)
Given this assumption, we can characterize the contents of the feedback utterances in
examples (1) and (3) as in (26) and (27):
(26) C1(ja, c1) = [s | & <<at l1; willing-to-continue, B1; 1>>,
<<at l1; perceive, B1, u1; 1>>,
<<at l1; understand, B1, u1, P1, 1>>,
<<at l1; accept, B1, P1; 1>> ]
Cs(ja, c1) = [s | & <<at l1; accept, B1; P1; 1>> ]
(27) C1 (ja de e de ju, , c3) = [s | & <<at l3; willing-to-continue, B3; 1>>,
<<at l3; perceive, B3, u2; 1>>,
<<at l3; understand, B3, U3, P3; 1>> ]
Cs (ja de e de ju, , c3) = [s | & <<at l3; accept, a3, P3; 1>>,
<<at l3; believe, a3, P3; 1>> ]
4.4 Context sensitivity with regard to the preceding communicative act
In this section we will discuss one kind of context sensitivity in relation to the
formalizations developed so far, namely sensitivity with respect factual polarity . In
section 3.4.3, we discussed two different analyses of the way in which the factual
polarity of the preceding utterance influences the content of words like yes and no,
one based on the notions of acceptance and rejection, one based on the notion of
commitment to facts. The formalizations suggested here are based on the first
analysis throughout.
In attribute-value notation, the occurrence content of nä/nej in examples (6) and (7)
can be represented as (28) and (29), respectively.
(6) A: de kan ju inte va för fiskarnas skull va
(it couldn't be for the sake of the fish you-know)
B: nä
(no)
24
(7) A så går naturen under me tekniken
(like that nature perishes with technology)
B: NEJ / / de växer upp annat då vet du
(no / / other things grow up you know)
(28) [INDICATE
[CONTACT +]
[PERCEPTION +]
[UNDERSTANDING +] ]
[SIGNAL
[ATITTUDE
[REJECT -] ] ]
(29) [INDICATE
[CONTACT +]
[PERCEPTION +]
[UNDERSTANDING +] ]
[SIGNAL
[ATTITUDE
[REJECT +] ] ]
(28) and (29) differ only in the value they assign to the PATH [SIGNAL [ATTITUDE
[REJECT]]]. In both cases, the value is the same as the polarity of the preceding
statement. We can capture this generalization in a set of constraints on the
attribute-value structure C representing the content of an utterance of the type nej in
the context of a preceding statement represented by the attribute-value structure PS
(where the notation f:path designates “the value assigned to path in feature structure
f”):
(30) C[INDICATE [CONTACT] ] = +
C[INDICATE [PERCEPTION] ] = +
C[INDICATE [UNDERSTANDING] ] = +
C[SIGNAL [ATTITUDE [REJECT] ] ] = PS[POLARITY]
Using situation semantics the contents of nä/nej in examples (6) and (7) can be
represented as in (3 1) and (32), where, for the first time, we try to give a little
structure to the contexts.
The context in example (6) is characterized as a situation c6 where it is the case that a
person A6 addresses B6 at a location l6-1 (temporally preceding the location l6 where
the feedback utterance occurs), making an utterance 146 with content P6, which has
the polarity 0.
25
In a similar way, the context of example (7) is characterized as a situation c7 where it
is the case that a person A7 addresses B7 at a location l7-1 (temporally preceding the
location l7 where the feedback utterance occurs), making an utterance u7 with content
P7, which has the polarity 1.
(31) C1 (nej, c6) = [s | & <<at l6; willing-to-continue, B6; 1>>,
<<at l6; perceive, B6, u6; 1>>,
<<at l6; understand, B6, u6, P6; 1>> ]
Cs(nej, c6) = [s | & <<at l6; reject, B6, P6; 0>> ]
where c6 | = <<at l6-1; address, A6, B6; 1>>
c6 | = <<at l6-1; utter, A6; u6; 1>>
c6 | = <<at l6-1; content, u6; P6; 1>>
c6 | = <<at l6-1; polarity-of, 0, P6; 1>>
(32) C1 (nej, c7) [s | & <<at l7; willing-to-continue, B7; 1 >>,
<<at l7; perceive, B7; u7; 1 >>,
<<at l7; understand, B7; u7; P7; 1 >> ]
Cs (nej, c7) [s | & <<at l7; reject, B7;, P7; l>> ]
where c7 | = <<at l7-1; address, A7, B7; 1>>
c7 | = <<at l7-1; utter, A7, u7; 1 >>
c7 | = <<at l7-1; content, u7; P7; 1 >>
c7 | = ~<at l7-1; polarity-of, 1, P7; 1 >>
We can generalize over (31) and (32) by means of parameters and obtain (33),
which is a characterization of the content of nej in a context of type c. (We use
boldface for parameters; note especially the polarity parameter I)
(33) C1 (nej, c) =[s | & <<at li; willing-to-continue, B; 1 >>,
<<at li; perceive, B, u; 1 >>,
<<at li; understand, B, u, P; l >> ]
Cs(nej, c) =[ s | & <<at li; reject, B, P; I>> ]
where c |= <<at li-1; address, A, B; l>>
c |= ~<at li-1; utter, A, u; l>>
c |= <<at li-1; content, u, P; 1>>
c |= <<at li-1; polarity-of, I, P; 1>>
26
Since we have not yet worked out a formalized way of capturing the systematic
dependency of feedback content on the speech act (mood) and information status of
the preceding utterance, we will not discuss formalization with regard to these
features of context sensitivity here.
4.5. Evocative function
Adding evocative functions to the representations used so far requires nothing new in
principle. For the attribute-value representations we simply add a complex-valued
feature EVOCATIVE, which takes as its value feature structures representing
different evocative functions. For the situation semantic approach we simply extend
the situation-types representing the content of feedback utterances with more
conditioning infons corresponding to the evocative aspects of the utterances.
However, since we have not yet worked out a precise and detailed account of the
evocative functions we also abstain from giving any formalized examples here.
5. Conclusions
In this paper, we have argued, discussed and at least partly demonstrated that what
we have called “the linguistic feedback system” of a language should not be
regarded as an area of hopeless complexity and confusion.
Rather, linguistic feedback systems seem to be describable subsystems of the
interactive mechanisms available in the spoken form of any language. We have
further argued, that such systems not only can be described phonologically,
morphologically and syntactically, but also semantically and pragmatically. In order
for such a description to be possible, feedback expressions and feedback mechanisms
must, just like deictic expressions and deictic mechanisms, be regarded as highly
context dependent. Specifically, we have argued that an account of the meaning of
feedback utterances involves considering at least the following dimensions of
content.
1 . Type of reaction to preceding communicative act.
2. Communicative status of various aspects of the content conveyed by the
feedback utterance.
3. Context sensitivity with regard to the preceding utterance in at least the
following respects:
(i) Its evocative function (type of speech act)
(ii) Its factual polarity
(iii) Its information status.
27
4. Evocative status of the feedback utterance.
Our attempts at formalizing these features of the meaning of feedback utterances
using attribute-value matrices and situation semantics are, naturally, only first
attempts but, we hope, sufficiently precise as to convince other linguists that the area
of feedback might be worthy of their attention.
28
References
Ahlsén E. 1985. Discourse Patterns in Aphasia. Gothenburg Monographs in
Linguistics, 5. University of Göteborg, Dept of Linguistics
Allwood, J. 1976. Linguistic Communication as Action and Cooperation.
Gothenburg Monographs in Linguistics, 2. University of Göteborg, Dept of
Linguistics.
Allwood, J. 1978. On the analysis of communicative action. Gothenburg Papers in
Theoretical Linguistics, 38. University of Göteborg, Dept of Linguistics.
Allwood, J. 1979. Ickeverbal kommunikation: en översikt. Papers in
Anthropological Linguistics, 2. University of Göteborg, Dept of Linguistics.
Allwood, J. 1981. On the distinctions between semantics and pragmatics. In W.
Klein and W.J.M. Levelt (eds) Crossing the Boundaries in Linguistics.
Dordrecht: Reidel.
Allwood, J. 1984. Finns det svenska kommunikationsmönster? In Vad är svensk
kultur? Papers in Anthropological Linguistics, 9. University of Göteborg,
Dept of Linguistics.
Allwood, J. 1985. Tvärkulturell kommunikation. In J. Allwood (ed) Tvärkulturell
Kommunikation Papers om Anthropological Linguistics, 12. University of
Göteborg, Dept of Linguistics.
Allwood, J. (ed) 1988a. Feedback in Adult Second Language Acquisition. Final
Report 11, Ecology of Adult Second Language Acquisition (ESF).
Allwood, J. 1988b. Om det svenska systemet för språklig återkoppling. In P. Linell,
V. Adelswärd, T. Nilsson, and P.A. Pettersson (eds) Svenskans beskrivning,
16, SIC 21 a. University of Linköping, Dept of Communication Studies.
Allwood, J. 1988c. Återkoppling i vuxnas språkinlärning. In K. Hyltenstam and I.
Lindberg (eds) Svenska som andraspråk, University of Stockholm, Centre for
Bilingualism Research.
Allwood, J. 1990. Synopsis of a method for conceptual determination. Paper
presented at the Twelfth Scandinavian Conference of Linguistics, Reykjavik.
29
Allwood, J., Nivre, J. and AhIsén, E. 1990. Speech management - on the non-written
life of speech. Nordic Journal of Linguistics, 13:1-48.
Anward, J. 1986. Emotive expressions. In Ö. Dahl (ed) Papers from the Ninth
Scandinavian Conference of Linguistics. University of Stockholm, Dept of
Linguistics.
Barwise, J. 1986a. The Situation in Logic - II: Conditionals and Conditional
Information. In E.C. Traugott, C.A. Ferguson and LS. Reilly (eds) On
Conditionals. Cambridge: Cambridge University Press.
Barwise, J. 1986b. On the Circumstantial Relation between Meaning and Content.
Versus 44:23-39.
Barwise, J. 1989. The Situation in Logic. CSLI Lecture Notes 17, Stanford: CSLI
Publications.
Barwise J. and Perry, J. 1983. Situations and Attitudes. Cambridge, Mass.: MIT
Press.
Barwise J. and Perry J. 1985. Shifting Situations and Shaken Attitudes. Linguistics
and Philosophy, 8:150-161.
Brown, G. and Yule, G. 1983. Discourse Analysis. Cambridge: Cambridge
University Press.
Clark, H. and Schaefer E.f. 1989. Contributing to Discourse. Cognitive Science, 13,
259-294.
Fries, C. 1952. The Structure of English. London: Longman.
Hellberg, S. 1985. Scandinavian sentence types. University of Göteborg, Dept of
Nordic languages.
Heritage, J. 1984. A change-of-state token and aspects of its sentential placement. In
J. M. Atkinson and J. Heritage (eds) Structures of Social Action. Cambridge:
Cambridge University Press.
Johnson, M. 1988. Attribute-value Logic and the Theory of Grammar. CSLI Lecture
Notes 14, Stanford: CSLI Publications.
Levinson, S. 1985. Pragmatics. Cambridge: Cambridge University Press.
30
Nivre, J. 1991. Feedback and Situation Theory. Gothenburg Papers in Theoretical
Linguistics 62. University of Göteborg, Dept of Linguistics.
Peirce, C.S. 1955. Philosophical writings of Peirce. Edited by J. Bülchler. New
York: Dover.
Israel, D. and Perry, J. 1988. What is Information? Paper presented at The
Conference on Minds, Brains and Language, Vancouver, B.C., February
1988.
Plunkett, K. and Strömqvist, S. 1990. The acquisition of Scandinavian languages.
Gothenburg Papers in Theoretical Linguistics, 59. University of Göteborg,
Dept of Linguistics.
Pollard, C. and Sag, 1. 1987. Information -based Syntax and Semantics. CSLI
Lecture Notes 13, Stanford:CSLI Publications.
Searle, J. 1969. Speech Acts. Cambridge: Cambridge University Press.
Severinson-Eklundh, K. 1986. Dialogue Processed in Computer-mediated
Communication, A study of letters in the COM-system, University of
Linköping, Dept of Communication Studies.
Schegloff, E. 1982. Discourse as an Interactional Achievement. In D. Tannen (ed)
Analyzing Discourse: Text and Talk. (Georgetown University Round Table
on Language and Linguistics, 1981). Washington, D.C.: Georgetown
University Press.
Shieber, S.M. 1986. An Introduction to Unification Based Approaches to Grammar.
CSLI Lecture Notes 4, Stanford:CSLI Publications.
Sigurd, B. 1984. Om Jaså, Bra, Precis och andra returord. Språkvård 3:1984,3-8.
Wiener, N. 1948/1961. Cybernetics; or Control and Communication in the Animal
and in the Machine. Cambridge, Mass.: MIT Press.
Yngve, V. On getting a word in edgewise. In Papers from the sixth Regional
Meeting, Chicago Linguistic Society, p. 567-577.
... The recent development of big data and deep learning techniques has greatly advanced both task-oriented and nonoriented dialogue systems, which has encouraged a huge amount of deep learning based researches in dialogue systems. In this article, we aim to (1) give an overview about dialogue systems especially recent advances from deep learning; and (2) discuss possible research directions. The remaining of the article is organized as follows. ...
... [43] simulated dialogues between two virtual agents. They defined simple heuristic approximations to rewards that characterize good conversations: good conversations are forward-looking [1] or interactive (a turn suggests a following turn), informative, and coherent. The parameters of an encoder-decoder RNN defined a policy over an infinite action space consisting of all possible utterances. ...
Preprint
Dialogue systems have attracted more and more attention. Recent advances on dialogue systems are overwhelmingly contributed by deep learning techniques, which have been employed to enhance a wide range of big data applications such as computer vision, natural language processing, and recommender systems. For dialogue systems, deep learning can leverage a massive amount of data to learn meaningful feature representations and response generation strategies, while requiring a minimum amount of hand-crafting. In this article, we give an overview to these recent advances on dialogue systems from various perspectives and discuss some possible research directions. In particular, we generally divide existing dialogue systems into task-oriented and non-task-oriented models, then detail how deep learning techniques help them with representative algorithms and finally discuss some appealing research directions that can bring the dialogue system research into a new frontier.
... According to Allwood et al. (1992), feedback can be illustrated as a ladder with four rungs: contact, perception, understanding, and attitudinal reactions. While attentiveness mainly deals with the lower levels of feedback (contact and perception), cooperativeness (C) represents the user's willingness to express understanding and attitude. ...
... Certain features are assumed to influence others: As shown in (Khawaja et al., 2014) and discussed earlier, expertise impacts the cognitive load of the user. At the same time, attentiveness is a requirement for cooperativeness (Allwood et al., 1992), which explains their dependency. ...
... Vocal backchannels like 'uh-huh', 'wow' and 'yeah' often serve the role of non-interrupting feedback in the background that the active speaker may or may not explicitly react to [7,8]. They constitute a subset of short vocal feedback responses including yes/no responses and agreements, which may or may not be regarded as a separate turn [9,10]. ...
... Like dialogue acts, feedback responses can be classified into various communicative functions [9,11,12]. Recent work defines different types of feedback functions, emphasizing their diverse roles and context-dependent nature in conversations [13,14]. Studies also suggest that speakers align their feedback prosodically with the preceding speech from the interlocutor, further underlining their context-dependence [15,16]. ...
... To verify the establishment of a Common Ground during a dialogic interaction, different linguistic or para-linguistic feedback analysis strategies (Traum, 1999) can be exploited. From a linguistic point of view, dialogue efficiency can rely on the analysis of communicative feedback, whose relevance was pointed out by Allwood et al. (1992) and which continues to be considered a fundamental characteristic in dialogue modelling (Buschmeier and Kopp, 2018). In this work, specific attention is dedicated to computational pragmatics, with respect to the systems' use of pragmatic tools in specific inconsistent contexts and their impact on the human capability to solve them. ...
Article
In linguistics, research on dialogue systems has accentuated the need to focus on various pragmatic aspects for their management and modelling. Among the most important pragma-linguistic speech acts in dialogue systems studies are Clarification Requests, corrective feedback that in some circumstances require access to the set of shared knowledge known as Common Ground. Regarding Common Ground management, pragmatic studies suggest differences in the type of polar questions that people prefer be used in Clarification Requests, where polar questions can have two possible answers: true or false. This preference appears to depend on the relationship between bias and contextual evidence. In this work, we show that varying the form of polar questions in a given pragmatic setting can influence the capability of people to track Common Ground inconsistencies. As a result, we demonstrate that using a negative polar question in Italian has functional consequences when communicating conflicting material in the Common Ground. This can improve the quality of human interactions with dialogue systems, in terms of an improved identification of the conflict. The results obtained in this work provide insights into design of error reporting approaches in natural interactions.
... B: No, I don't want to. (Horn's 33) The assumption seems to have been that A can identify these forms as rejections by recognizing logical inconsistency either directly from what was said, or via an inferential chain [Gazdar, 1979, Allwood, 1992 inter alia. Thus much of previous work would lead one to believe that there is a neat trichotomy with the processing that A must do to monitor additions to the common ground: 1. ...
Preprint
This paper discusses the processes by which conversants in a dialogue can infer whether their assertions and proposals have been accepted or rejected by their conversational partners. It expands on previous work by showing that logical consistency is a necessary indicator of acceptance, but that it is not sufficient, and that logical inconsistency is sufficient as an indicator of rejection, but it is not necessary. I show how conversants can use information structure and prosody as well as logical reasoning in distinguishing between acceptances and logically consistent rejections, and relate this work to previous work on implicature and default reasoning by introducing three new classes of rejection: {\sc implicature rejections}, {\sc epistemic rejections} and {\sc deliberation rejections}. I show how these rejections are inferred as a result of default inferences, which, by other analyses, would have been blocked by the context. In order to account for these facts, I propose a model of the common ground that allows these default inferences to go through, and show how the model, originally proposed to account for the various forms of acceptance, can also model all types of rejection.
... response corresponds roughly to the backward looking acts of Allen and Core (1997) and several dimensions of the ISO scheme. All of these relations provide some sort of feedback (Allwood et al., 1992), as to how the responder has perceived, understood, and reacted to the antecedent. The feedback can be positive or negative at multiple levels. ...
Preprint
In this paper, we describe the development of symbolic representations annotated on human-robot dialogue data to make dimensions of meaning accessible to autonomous systems participating in collaborative, natural language dialogue, and to enable common ground with human partners. A particular challenge for establishing common ground arises in remote dialogue (occurring in disaster relief or search-and-rescue tasks), where a human and robot are engaged in a joint navigation and exploration task of an unfamiliar environment, but where the robot cannot immediately share high quality visual information due to limited communication constraints. Engaging in a dialogue provides an effective way to communicate, while on-demand or lower-quality visual information can be supplemented for establishing common ground. Within this paradigm, we capture propositional semantics and the illocutionary force of a single utterance within the dialogue through our Dialogue-AMR annotation, an augmentation of Abstract Meaning Representation. We then capture patterns in how different utterances within and across speaker floors relate to one another in our development of a multi-floor Dialogue Structure annotation schema. Finally, we begin to annotate and analyze the ways in which the visual modalities provide contextual information to the dialogue for overcoming disparities in the collaborators' understanding of the environment. We conclude by discussing the use-cases, architectures, and systems we have implemented from our annotations that enable physical robots to autonomously engage with humans in bi-directional dialogue and navigation.
... response corresponds roughly to the backward looking acts of Allen and Core Draft (1997) and several dimensions of the ISO scheme. All of these relations provide some sort of feedback (Allwood et al., 1992), as to how the responder has perceived, understood, and reacted to the antecedent. The feedback can be positive or negative at multiple levels. ...
Article
Full-text available
In this paper, we describe the development of symbolic representations annotated on human–robot dialogue data to make dimensions of meaning accessible to autonomous systems participating in collaborative, natural language dialogue, and to enable common ground with human partners. A particular challenge for establishing common ground arises in remote dialogue (occurring in disaster relief or search-and-rescue tasks), where a human and robot are engaged in a joint navigation and exploration task of an unfamiliar environment, but where the robot cannot immediately share high quality visual information due to limited communication constraints. Engaging in a dialogue provides an effective way to communicate, while on-demand or lower-quality visual information can be supplemented for establishing common ground. Within this paradigm, we capture propositional semantics and the illocutionary force of a single utterance within the dialogue through our Dialogue-AMR annotation, an augmentation of Abstract Meaning Representation. We then capture patterns in how different utterances within and across speaker floors relate to one another in our development of a multi-floor Dialogue Structure annotation schema. Finally, we begin to annotate and analyze the ways in which the visual modalities provide contextual information to the dialogue for overcoming disparities in the collaborators’ understanding of the environment. We conclude by discussing the use-cases, architectures, and systems we have implemented from our annotations that enable physical robots to autonomously engage with humans in bi-directional dialogue and navigation.
... During conversations, listeners produce vocal, visual and multimodal responses or reactions known as feedback, which serve as explicit markers of attention, interest and understanding (Allwood et al., 1992;Bunt, 2012;Schegloff, 1982) and guide conversational flow (Bertrand, 2021;Gandolfi et al., 2023). ...
Article
This article presents a different experiment examining the impact of feedback timing on its perception. Dialog sequences, featuring a main speaker's utterance followed by a listener's feedback, were extracted from spontaneous conversations. The original feedback instances were manipulated to be produced earlier, up to 1.5 s in advance, or to be delayed, up to 2 s later. Participants evaluated the feedback acceptability and engagement level of the listener. The findings reveal that 76% of the time feedback remains acceptable regardless of the delay. However, engagement decreases after a 1-s delay while no consistent effect is observed for feedback anticipation.
Article
Full-text available
1. Inledning 1.1 Terminologi Tvärkulturell kommunikation eller kommunikation mellan människor med olika kulturell bakgrund har alltid varit och kommer säkerligen att förbli en viktig förutsättning för mänsklig samexistens på jorden. Denna uppsats syftar till att visa hur man inom ramen för en allmän modell för mänsklig, främst språklig, kommunikation även kan ge en redogörelse för de faktorer som är viktiga i tvärkulturell kommunikation. Termen tvärkulturell har valts snarare än de i stort sett synonyma termerna interkulturell och korskulturell därför att termen ansluter till språkbruk som tvärvetenskaplig d v s samverkan mellan människor med olika vetenskaplig bakgrund. Kanske har termen också en något mindre förtingligande innebörd än interkulturell. Det är inte kulturer som kommunicerar, vad detta nu skulle kunna innebära, utan människor med olika kulturell bakgrund. Termen korskulturell har valts bort eftersom den av flera andra skäl inte går att tolka på något lättbegripligt sätt på svenska.
Book
Because of the ease of their implementations, attribute-value nased theorires of grammar are becoming increasingly populaar in theoretical linguistics as an alternative to transformational accounts, as well as in computational linguistics. Mark Johnson provides a formal analysis of attribute-value structures, of their use in a theory of grammar, of the representation of grammatical relations in such theories of grammar, and the implications of different representations. A classical treatment of disjunction and negation is alo included. "Essential reading for anyone interested in recent unification-based approcahes to grammar. Johnson lucidly lays out a formal framework in which a sharp distinction is drawn between descriptions of linguistic objects and the objects themselves. Negation and disjuntion over complex features, though linguistically desirable, have given rise to many problems, and one of Johnson's main achievements is to show that they can be interpreted using classic logic." -Ewan Klein, University of Edinburgh MARK JOHNSON is assitant professor of cognitive and linguistic sciences at Brown University.