From Affect to Grammar: Ritualization of Facial Affect
in
Signed Languages
James MacFarlane
University of New Mexico

0.  Introduction
 
 All definitions of grammaticization involve the change of a lexical item to a grammatical item.  Consider the following definitions:

 (1) "The attribution of grammatical character to an erstwhile autonomous   word."  (Meillet 1958[1912]: 131)

 (2) "A process which turns lexemes into grammatical formatives and makes grammatical formatives still more grammatical." (Traugott 1997:2)

 (3) "The process by which a frequently used sequence of words or morphemes becomes automated as a single processing unit." (Bybee,1997)

These definitions pose a problem for signed languages because many of the grammatical features found in signed languages seem to have developed from gesture rather than from lexemes.  Thus in American Sign Language (ASL), the majority of grammatical inflections have most likely evolved directly from non-lexical gestural sources, bypassing the lexicon altogether (Petitto 1991).  This appears to be the case in other signed languages as well.  These gestural, non-lexical origins can be found in two domains: manual gestures and non-manual gestures.  In this study I will focus on non-manual gestures, specifically raised eyebrows, in fifteen natural signed languages.  It is argued that these non-lexical/gestural sources have grammaticized in natural signed languages and continue to grammaticize.  This observation raises some important theoretical questions for grammaticization theory.  Do grammatical constructions (grams) that evolve from gestural/non-lexical sources follow the source determination hypothesis (the actual meaning of the construction that enters into grammaticization uniquely determines the path that grammaticization follows and, consequently, the resulting grammatical meanings)?  Do grammatical morphemes that develop out of gestural/non-lexical constructions also produce predictable, unidirectional semantic change?  Do grammaticizations that derive from gestural/non-lexical sources follow cross-linguistically similar paths for the development of grammatical meaning?  Since semantic substance evolves in grammaticization and the meaning of the source construction determines the subsequent grammatical meaning, do these gestural/non-lexical sources have an innate meaning similar to the lexical sources used in spoken languages?

1. The Data

The data for this study was taken from fifteen natural signed languages. Table 1.1 provides a characterization of the use of raised eyebrows in the fifteen languages in my sample for four grammatical markers: yes/no questions, topic, protasis and connectives which are Wh-word signs such as why, who and how.  The majority of the literature on Wh-words in this context has labeled them as "rhetorical questions" (e.g. Baker & Cokely 1980, Valli & Lucas 1992).  A more accurate analysis of these would take into consideration that in ASL they appear to be less of an indirect speech act, and instead function to link one clause to the next (Janzen, 1998).  I will discuss this type of connective in more detail later.

The sampling method used for this project employed a questionnaire methodology.  The results of the questionnaire were tabulated and then controlled for genetic bias. For several of the languages informants could not state definitively whether or not signers employed raised eyebrows to grammatically mark a specific category. In these cases, the category is marked with an "?" in the table.  It is important to note, however, that of the languages sampled, only Langue du Signe Quebecois (LSQ) does not use raised eyebrows to grammatically signal  one or more of the four grammatical categories (Dubuisson, Boulanger, Desrosiers & Lelievre 1991). Nonetheless, raised eyebrows (in LSQ); have been noted to accompany some of the grammatical categories mentioned above. A sampling of these grammatical markers in ASL is given below.
 
        ___________________ y/n
  (4) JOHN BECOME ANGRY?
          'Is John angry?'   (Baker & Cokely 1980:123)

Any type of question that seeks a positive or negative response is marked with raised eyebrows over the entire clause and has the whole clause under its scope.  Its obligatoriness is apparent because there is no other marker in the clause that marks it as a question.  Without the raising of the eyebrows, (4) would be understood as a statement that "John got angry". Topics are marked with the same marker that yes/no questions employ, except that sometimes the head is tilted slightly backwards.  It is clear that these constructions are not polar questions because the raised eyebrows do not extend over the entire clause.

        __________________ t
 (5) YESTERDAY MEETING, BORING.
  'The meeting yesterday was really boring.' (Baker & Cokely 1980:159)
 
Wh-word signs that behave like connectives are often translated as 'because'. In addition to the lexical item, eyebrows must be raised and frequently the head is tilted (Baker & Cokely 1980:137).

                    __t__                                                                           _t_
 (6) POSS.1 GOOD.FRIEND PRO.3 SAY THERE ALCOHOL IN WHY   POSS.3 MOTHER PREVIOUS WORK ALCOHOL STORE (...)
  'My friend says there's alcohol in it (the punch) because her mother used to   work in a liquor store, (and...)' (Janzen, 1998.)

Thus, when a Wh-word is used in this kind of context, rather than functioning as any kind of question, it is simply marking the first clause as a kind of topic (Janzen 1998 & Herring 1991:265).  In a sense, these constructions derive from rhetorical questions. But due to the fact that they function to link two clauses these constructions would be more appropriately described as connectives .  This grammaticization of rhetoricals has been treated by others as well (cf. Hopper 1991)  Thus the rhetorical questions in my data will be treated as connectives.  Finally, raised eyebrows mark the protasis of conditional statements.
 

      ______________cond    (gaze right     )  nodding
 (7) TOMORROW SNOW, PRO.1 GO-TO-rt SKI.
  'If it snows tomorrow,  I'm gonna go skiing.' (Baker & Cokely 1980:143)

2.  Gesture to Grammar
The idea that human language began its evolution as an iconic visual-gestural system is not new.  Givon (1995), Armstrong, Stokoe & Wilcox (1995) & Hewes, (1974) have all supported this hypothesis.  The grammatical code is probably the most recent evolutionary addition to human communication, this hypothesis is supported by two very strong arguments.  First, both hearing and signing children acquire the lexicon much earlier than the grammar (Reilly, McIntire & Bellugi 1990).  Second, birds, dogs, horses, primates and other pre-human species are easily taught auditory or visual lexical code-labels for lexical items such as nouns and verbs, but are unable to acquire grammar (Givon, 1995: 400).  Givon further states that "these more abstract elements of the grammatical code are probably second-order constructions, inferred from the more concrete signals."  Thus when grammar is developed it codes both propositional-semantic information (clauses) and discourse-pragmatic coherence (discourse) (Givon, 1995: 401).

I suggest that raised eyebrows in signed languages evolved from a signal, which was frequently used with a specific communicative behavior, that later lead to ritualization.  That is, raised eyebrows were emancipated from their primary motivation  and made free to serve a communicative function instead.  As a result of this emancipation the signal is free to undergo automation, signal-reduction, abstraction and schematicization (Haiman 1994).  Givon states that this is as true in the development of native human sign languages as it is in animal communication. How did this gesture become emancipated from its original function?  I follow Janzen's (1998) proposal that raised eyebrows, head tilted forward and eyes opened slightly wider is a common human gesture of anticipation and encouragement of a response.  This gesture can be conceived of in terms of Eckman's (1971: 216) Facial Affect Program (c.f. Reilly, McIntire & Bellugi 1990).

Thus an emotion elicited by some event, the nature of that event typically varying with culture, activates the facial affect program.  This program links each primary emotion to a distinctive patterned set of neural impulses to the facial muscles.  When anger is elicited, one set of muscular movements will be triggered; when interest is elicited, a different set of muscle movements will be triggered. Eckman claims that this program is constant for all human beings.  What is universal in facial expressions of emotion is the particular set of facial movements triggered when a given emotion is elicited.  It is therefore reasonable to conclude that when the emotion of interest or anticipation of a response is felt, humans automatically trigger the facial affect program that is linked to that specific emotion.  As Eckman puts it, "The activation of this affect program can (italics are original) result in a characteristic facial appearance for fear, anger, sadness, disgust, surprise, happiness and interest for all human beings."
 
What role, if any, do cultural influences play in governing these facial expressions?  Eckman introduces the concept of display rules to describe a cultural mechanism which can override the affect program and control facial appearance.  Therefore, the elicitors of the emotions, the display rules which govern facial appearance, and the behavioral consequences of emotion are all shaped by culture.
 This well-coded signal of interest or anticipation of a response, then, is present in the facial affect program in all humans.  Somewhere along the way, either during the evolution of languages in general or during the natural development of signed languages, raised eyebrows, a tilted forward head and widening of the eyes was emancipated from its primary motivation. In turn, this gesture became free to serve a communicative function linked with the emotion at hand.  As Haiman (1994: 5) puts it, "In this sense, ritualization is the acquisition of meaning.  A ritual is identified when it ceases to be a purely instrumental act and becomes a sign."

3.  The grammaticization path
 Janzen (1998) has proposed a grammaticization pathway for topics in ASL.  This path is given in (8):
 
 (8)    communicative questioning gesture > yes/no questions > topic constituents > connectives > conditionals > temporal reference shift.
I will assume this grammaticization pathway for this study, with one exception: I will call the communicative questioning gesture interest/anticipation facial affect program (IFAP).  An overview of the grammaticization pathway proposed by Janzen is given below. Additionally I will not discuss temporal reference shift here as I have not included this in my sample. Thus, the grammaticization path I am proposing is represented in (9).
 
(9)  IFAP >  Y/N-Q   >  TOPIC   >  CONNECTIVE  >    PROTASES
 
I will begin with yes/no questions as I have already addressed the ritualization/grammaticization of the IFAP.

3.1  Yes/no questions > topic
 Janzen states that "topics essentially take the form of yes/no questions, but for several important differences.  For one, whereas yes/no questions can take the form Are you doing X?, topics imply the question Do you know X?."  The difference between these two is that yes/no questions ask the addressee to respond about situations external to the linguistic situation and his or her participation in them.  Topics, however, ask the addressee if the situation reflected in the topic is within his or her immediate knowledge base (Janzen 1998).  This is reflected in BSL as well:
         ___________ t____
 (10) HOUSE THERE BIG, MY FATHER BEEN BUILT.
  "That big house over there, my father built it."     (Allsop, L. 1988: 28:20)

Furthermore, the addressee is not expected to respond to the topic-marked constituent as if it were a real question.  Janzen suggests that the mechanism of change for yes/no questions to topic follows the semantic-pragmatic change suggested by Traugott (1989:31).
 
(11) propositional > ((textual) > (expressive))

Thus in the context of yes/no questions in ASL the whole construction of the proposition and the questioning of its truth value are the source for grammaticization (Janzen 1998).

3.2  Topic > Conditionals and Connectives
 As outlined above, the yes/no marking has abstracted away from the more direct communicative function that yes/no questions have, but retain the yes/no marking (Janzen 1998).  Janzen proposed that a further development along the grammaticization path is linking pieces of text rather than organizing information from the shared experience of the interlocutors, namely conditionals and connectives.  Thus the example in (12) from LIS shows how the conditional construction has taken on a discourse function which is textual.
 
        _cond_
 (12) PIOVE IO CINEMA ANDARE NO.
         Rain      I    Cinema        go          no
      "Se piove io non vado al cinema."  (Franchi, M.L., 1987:170)
      "If it rains, I won't go to the cinema."

Similarly, in SSL (13), what is described as a rhetorical question might be better described as a kind of topic where the Wh-word vat has been emancipated form the interactiveness of a question form and adopted topic marking (cf. Janzen 1998).  The context for the example in (13) is a deaf person telling about school and a teacher who is impossible to lip-read.

          __cond___________________   _rhet_.
 (13) O-M JAG KOMMA -DIT-c-m ,  VAD La_RA-SIG MIG, VAD ......
          cond                        _rhet__.
          I-F I GO-THERE-r-m      WHAT LEARN-ME,  WHAT.......
          "If I go there, what can I learn, what........" (Bergman, B. 1984:57)
 
4.  Grammaticization theory
 Because this study is both exploratory and hypothesis-oriented it is necessary to test whether the grammaticization pathway given here is applicable to a set of hypotheses for a theory of grammaticization.  Due to the claim that the grammaticization path above developed out of gesture, it would be useful to know if these grams behave similarly to those that were grammaticized from lexical sources.  I will address several of the hypotheses proposed for grammaticization theory in what follows.

4.1 Source Determination
 The source determination hypothesis states that the actual meaning of the construction that enters into grammaticization uniquely determines the path that grammaticization follows and, consequently, the resulting grammatical meanings.  It has been noted that what enters into grammaticization is the reference plane of basic, irreducible notions which serves as the basis for grammatical meaning in human languages (Bybee et al. 1994: 10).  Additionally, Heine, Claudi and Hunnemeyer (1991) state that:
 

source concepts may be said to refer to some of the most elementary human experiences; they are typically derived from the physical state, behavior, or physical environment of man and are frequently referred to in thought and communication.  They are likely to form part of 'the alphabet of human thoughts' and belong to a universal set of semantic parameters.

These notions are also culturally independent in that "they tend to be conceived of in a similar way across linguistic and ethnic boundaries" (Heine et al. 1991: 33).

The IFAP does have its roots in an extremely elementary human experience and seems to belong to that universal set of semantic parameters that Heine et al. speak of.  From these basic, irreducible notions, units enter into grammaticization and mark some grammatical category that is linked with the unit's inherent meaning.  For example, constructions involving spatial relations generally employ body part terms (Svorou 1986, 1993 & Heine et al., 1991) such as the English phrase The back of the paper.  Thus the lexical source, the human back, uniquely determined the path that grammaticization follows.

 The source determination hypothesis applies to the grammaticization path proposed in (9) as well. The evidence suggests that the source for these grammatical categories is the basic human notion of interest.  This emotion is a basic human experience that can be shared by all humans.  Therefore, the IFAP would not be an unlikely source for the development of the grammaticization pathway proposed in (9), particularly for those persons who rely on visual/gestural forms of communication.

4.2  Unidirectionality
 Cross-linguistic study has revealed that grams will evolve from some concrete substance to a more general and abstract grammatical meaning (Bybee et al. 1994).  It has been observed that these grams evolve unidirectionally; that is, grams do not move in the reverse direction of a grammatical pathway (Bybee et al. 1994; Hopper & Traugott 1993).
The grammaticization pathway proposed in (9) is also viewed as unidirectional.  We see that the IFAP evolves to yes/no questions, which may then evolve into topic constituents and so on, but the reverse direction is unknown.  Thus if raised eybrows are used to mark the protasis, raised eyebrows will also signal  yes/no questions.  This prediction is borne out by my data.

 In addition to the unidirectionality of semantic change, there is some evidence for unidirectionality in the grammatical and phonological change that accompanies the grammaticization path proposed in (9).  For example, Baker & Cokely (1980) note that Yes/No questions in ASL are accompanied by a brow raise, widened eyes,  and frequently, a forward tilting of the head or body.  This implies phonological reduction, because forward tilting of the head is not obligatory.  The same reduction for yes/no questions is seen in SSL.  Bergman (1984: 53) notes that a yes/no question can also be signaled completely non-manually by only raising the eyebrows or slightly tilting the head with no accompanying signs.  For signaling protasis, Bergman states that "I have found in Swedish Sign Language no example where head and body are tilted forwards,"  In ASL, however, the grammatical marker of the conditional clause is lifted eyebrows combined with a tilting of the head and, in some cases, the tilting of the body in the same direction as the head tilt (Baker & Cokely 1980: 141).  Finally, more phonological reduction is seen in the topic marking of Wh-word connectives in LIS.  Celo (1995) states that a single fast movement of the eyebrows signal "rhetorical questions" in LIS.  This gram is greatly reduced from the original IFAP gesture of raised eyebrows, head tilted slightly forward, and eyes opened slightly wider.  This is an indication that phonological reduction accompanies the unidirectional grammaticization path proposed in (9).

 It is also important to note that there is nothing deterministic about grammaticization and unidirectionality (Hopper & Traugott, 1993: 95).  Changes do not have to occur, nor do they have to complete the grammaticization pathway.  That is, a particular grammaticization process (in a particular language)  may be arrested before it reaches the end of the path (Hopper & Traugott, 1993:95).  Thus, it appears that once gesture becomes ritualized and acquires grammatical meaning, it must move unidirectionally just as any other gram in any other language.

4.3  Universal Paths
 Bybee et al. state that "any grammaticizations that begin with the same or similar source meaning can be expected to follow the same course of change."  As a result, there will be some cross-linguistically similar paths for the development of grammatical meaning.  However, some language-specific instances of grammaticization can arise by idiosyncratic selection of source material.  It is my claim that the source material for the grammaticization pathway proposed in (9) is universal among signed languages, but just as with spoken languages, signed languages can take several possible paths.  This appears to be the case with LSQ.  LSQ, a naturally developed sign language in Quebec, does not use raised eyebrows to signal any of the categories noted above.  Dubuisson, Boulanger, Desrosiers & Lelievre (1991) state that "eyebrow movement is neither obligatory or meaningful in LSQ."  Although raised eyebrows have been noted to occur during the production of the four grammatical categories discussed in this paper, they are not grammatical.  Thus LSQ has grammaticized a different source to mark yes/no questions and the like.  The remaining languages in my sample do employ raised eyebrows to grammatically mark yes/no questions.

 If grammaticization paths are universal, can we expect to find the same path proposed in (9) for spoken languages?  Haiman (1978) discussed a universal grammaticization path for conditionals and topics that is presented in (14).

 (14) interrogatives >  topic  > conditionals

In this way the paths are universal.  Both signed and spoken languages use as their source some construction that implies interest or questioning.  Bybee et al. attribute the fact that certain grammaticization paths are common in diverse genetic and areal groups to the existence of common cognitive and communicative patterns underlying the use of language.  Thus it is reasonable to conclude that deaf people and hearing people share these common patterns and necessarily reflect those in their languages.

4.4  Retention of earlier meaning
 One of the basic claims about grammaticization is that semantic substance evolves.   Thus we should find that certain semantic nuances of the source construction  can be retained in certain contexts long after the grammaticization has begun (Bybee, et al. 1994).  If the original IFAP is elicited by interest or anticipation of a response we would expect to see certain semantic nuances of interest or anticipation reflected in subsequent constructions.  Obviously yes/no question are elicited by interest.  An example of this can be seen in the SSL example in (15).
 
        ___________y/n-q
 (15) Do_V/ STUM DU?
      "Are you deaf?"   (Bergman, B. 1984:53)

Topic constituents can also be viewed as questions in that they ask the addressee if the situation reflected in the topic is within his or her immediate knowledge base, or is a referent that is immediately available (Janzen, 1998).  Secondly, the addressee is not expected to respond directly to the topic constituent as if it were a true question.  Thus the semantics of topic constituents do involve establishment of agreement, and the speaker must solicit this agreement from the addressee.  Haiman (1978) notes that this is why contrastive topics are marked as questions: the question is a request for affirmation or recognition of existence of the topic.  Thus the SSL topic marked constituent in (16) can be shown to have questioning or interest as part of its semantics.
 
         ____t
 (16) P-E-R,  VARA-EXPERT-PA RITA.
         ____t
         P-E-R,  BE-EXPERT-AT DRAWING.
          "As for Per,  he's an expert at drawing." (Bergman, B. 1984:57)

In the same way that topic constituents can be interpreted as polar questions, so can protases.  Haiman (1978:570) states that "Not only do conditional statements resemble polar questions, but the reverse is also frequently true."  Jesperson (1940:347) suggested that conditionals were questions with implied positive answers as in (17).
 
(17) Are you sick? (yes)  Well then, stay behind.

This questioning is reflected in the SSL conditional statement in (18).  This is particularly question-like in view of the fact that there is no lexical conditional marker.
 
           __cond_
 (18)   DU SJUK, STANNA-KVAR
          ____cond_
          YOU SICK, STAY-BEHIND
          "If you are sick, stay behind."

Finally, the semantics of connectives have questioning as part of their semantics as well.
When analyzed as a type of topic as proposed by Herring (1991) this construction has interest as part of the grammatical meaning as well.  Under this analysis it becomes apparent that grammatical categories which have visual/gestural sources will retain certain semantic nuances of the source meaning.

4.5  Semantic & phonological reduction
 Bybee et al. (1994) argue that the semantic evolution of grams, from their sources to full maturity, is characterizable in terms of successive instances of generalization or semantic reduction.  In addition this semantic reduction is paralleled by phonetic reduction.  I have already argued in section 4.2 that the source that enters into the grammaticization pathway proposed in (9) undergoes some phonetic reduction .  While the gram in question has not undergone extreme phonetic erosion, it is true that some phonetic features such as tilting of the head forward and widening of the eyes have eroded. Also, with the LIS example of Wh-marked connectives, the temporal length of the marker has been greatly reduced.  This lack of phonetic reduction is likely due to the fact that the gram is not phonetically complex.  The main gesture, that of raising the eyebrows, consists of only one gesture and it appears to be the most entrenched.  The other gestures that have been noted, forward tilting of the head, widening of the eyes, are in the process of phonetic erosion.

 Semantic reduction of the gram was discussed in section 4.4.  This semantic reduction starts with pragmatic inferencing of the IFAP so that it can be used to grammatically signal yes/no questions.  As a result of semantic reduction, yes/no questions are then able to signal topic constituents and then with more reduction they can signal protases and connectives.  These conclusions suggest that grammaticizations that have their source in gesture behave similarly to grams that develop from lexical sources in that they both suggest a link between semantic and phonetic reduction in the evolution of grammatical material.

4.6  Layering
 Layering is the synchronic result of successive grammaticization of forms which contribute to the same domain (Hopper & Traugott, 1993:124).  This can be exemplified by the three futures in English, will, shall, and going to (Bybee et al. 1994:21).  This is not unusual to find an array of grammaticizing constructions of different sources sharing or competing for overlapping territories. I was unable to find any examples of layering in my data.  This may be due to the fact that the signed languages in my data are all relatively young.  Most national sign languages are only developed as a result of the advent of national education for the deaf.  This allows deaf people from all over the country to come together and in doing so, they standardize the language.  Another, perhaps more plausible, explanation for the lack of layering in my data is that I only looked for and requested the grammatical use of raised eyebrows.  Thus, if there are other competing grams within a particular language in my sample, I may not have found them.  LSQ, on the other hand, may have layering of grammatical material as it does use lexical items to signal yes/no question and the like.  Although Dubuisson, Boulanger, Desrosiers & Lelievre (1991) state that "eyebrow movement is neither obligatory or meaningful in LSQ," it has been noted that the eyebrows do raise during these constructions.  Thus it could be possible that the two constructions are competing for overlapping territories.  My intuition is that grams from gestural sources can and do allow layering but does not appear in my data.  More research is needed to determine whether these grams fulfill the grammaticization hypothesis that new layers are continually emerging.

5.  Conclusion
 In this paper I have shown that non-lexical gestural sources grammaticize in signed languages.  In addition these grammaticization paths do not stray from hypotheses about grammaticization in spoken language.  Indeed grammatical categories that have evolved from gestural sources exhibit source determination of the gram, unidirectional pathways, universal paths, retention of earlier meaning and most likely layering.  Thus this analysis requires a re-thinking of the definitions of grammaticization that includes this common source.

 This analysis also provides support for the argument of a bodily basis for the origin and evolution of language (e.g., Armstrong et al., 1995; Kimura, 1993; Arbib & Rizzolatti, 1996).  I have argued that one of the sources for language grounded in the body is Eckman's Facial Affect Program.  Indeed the literature shows theoretical support for the presence of innate motor programs such as Eckman's Facial Affect Program.  Rizzolatti and Gentilucci (1988) and Arbib and Rizzolatti (1996) discuss a vocabulary of motor acts coded at the single neuron level.  Thus the identification of gestural source concepts for universal grammaticization paths provides further evidence that language is not modular; rather, it is rooted in our gestural system.
 

References