Portal de Periódicos da CAPES

The Unity of Focus: Evidence from Sign Language (ASL and LSF)

2016; The MIT Press; Volume: 47; Issue: 2 Linguagem: Inglês

10.1162/ling_a_00215

1530-9150

Philippe Schlenker, Valentina Aristodemo, Ludovic Ducasse, Jonathan Lamberton, Mirko Santoro,

Language, Metaphor, and Cognition

In spoken languages, focus (i) is normally realized by phonological prominence, which in English is effected by higher pitch, greater loudness, and longer duration (e.g., Katz and Selkirk 2011). Semantically, it (ii) signals the activation of alternatives (e.g., Rooth 1996) and (iii) has diverse effects, ranging from contrastive (as in (1a)) to exhaustive (as in (1b)).1Finally, it has been speculated that (iv) the realization of focus is driven by a biological ''effort code'' whereby greater pitch excursions (and thus greater effort on the speaker's part) are associated with greater emphasis/importance (Gussenhoven 2001, 2004).2Following Wilbur (2012), as well as Crasborn and Van der Kooij (2013) and Kimmelman (2014), we argue that versions of all four properties hold of focus in American Sign Language (ASL) and Langue des Signes Français (LSF), which suggests that focus has a unified semantics and to some extent a unified semantics/phonology interface across modalities; in particular, contrastive and exhaustive focus can be realized by the same prosodic means. Earlier studies emphasized the diversity of focus realization in ASL and the importance of syntactic movement (Wilbur 2012), which made it hard to isolate the role of prosody. By contrast, detailed production studies of Sign Language of the Netherlands (NGT) and Russian Sign Language (RSL) displayed in-situ strategies of (corrective and informational) focus marking involving modulations of sign speed, size, and duration, combined with nonmanual markers in NGT (Crasborn and Van der Kooij 2013, Kimmelman 2014; see also Kimmelman and Pfau to appear). We complement Wilbur's findings by studying particularly simple paradigms in which movement is inapplicable, showing that in these cases ASL and LSF sign modulations and nonmanuals may also suffice to mark focus, with diverse semantic effects, ranging from contrastive to exhaustive, as in spoken language. Following in part Crasborn and Van der Kooij (2013), we suggest that a version of the ''effort code'' is arguably at work in ASL and LSF, focused signs being realized not just with raised eyebrows (as is standard), but also with greater amplitude, speed acceleration, and longer hold times, among others. Since ASL and LSF are both descended from Old French Sign Language, some important similarities between them are of course to be expected; but we still note some differences in our data, particularly the presence of forward leans in ASL but not in LSF (which, however, includes head nods).Following the typology of Vallduví (1992), Wilbur (1999) establishes that in ASL ''the primary indicator of stress marking is the significant increase in peak velocity of prominent signs'' (p. 236)3,and that as in English ''a single prominence is assigned to the right-most lexical item in the phrase'' (p. 237), but that (unlike in English) prominence cannot be moved, with the result that focus is preferably realized by movement; in Wilbur's (and Vallduví's) terms, with respect to focus realization ASL is ''[– plastic],'' unlike English but like Catalan. In contrast to Wilbur's analysis of ASL, in production experiments Crasborn and Van der Kooij (2013) (studying NGT) and Kimmelman (2014) (studying NGT and RSL) emphasized the role of non-movement-based strategies, in particular of prosody. In both studies, focus appeared in question-answer pairs or in corrective contexts (as in The woman is eating chocolate, right? – The woman is eating ice creamF (Kimmelman 2014) or Is your car red? – No, my car is blueF (Crasborn and Van der Kooij 2013)). In Kimmelman's summary of his comparative data (p. 130), RSL focus primarily involves manual prosody, with modulations of size, speed, and duration, as well as repetitions. NGT also uses these strategies (with different frequencies), but in addition it makes use of nonmanual prosody, involving eyebrow raising, backward head tilt, and head nod.While Wilbur emphasizes the diversity of strategies of focus realization in ASL, we complement her observations by showing that in a subset of ASL and LSF cases in which movement is inapplicable, prominence and nonmanuals alone—including forward leans—can realize both contrastive and exhaustive focus, which makes for a particularly minimal comparison with English (some cases of in-situ focus also appear in Wilbur and Patschke 1998 and Lillo-Martin and Quadros 2008). We selected two cases of focus that are semantically very different: instances of contrastive focus that do not involve association with any operator and that could be analyzed in terms of simple constraints on givenness (Schwarzschild 1999); and instances of embedded focused elements that trigger exhaustive readings in the scope of other operators and that would require, in the analysis of Chierchia, Fox, and Spector(2012), the insertion of exhaustivity operators—with clear truth-conditional consequences. Importantly, we will show that despite these semantic differences, these two types of focus can be realized with the same prosodic means.Our data were elicited from Lamberton for ASL and from Ducasse for LSF. Both are Deaf children of Deaf, signing parents, and they co-authored this squib. In contrast to the production experiments reported in Crasborn and Van der Kooij 2013 and Kimmelman 2014, we obtained grammaticality judgments using the ''playback'' method, with repeated quantitative acceptability judgments (1–7, with 7 = best) and repeated inferential judgments (on separate days) on videos involving minimal pairs (for earlier uses of this method, see Schlenker 2013, 2014, to appear, Schlenker, Lamberton, and Santoro 2013, Kuhn to appear). We asked each consultant to put emphasis on certain words, and thus the production part was artificial, just as it would be for the creation of experimental stimuli; but the subsequent (and repeated) judgment task (by the same consultant) made for minimal comparisons between appropriate and deviant sentences depending on the placement of focus marking. Like the standard ''introspective'' method used in linguistics, our procedure provides some data that are hard or impossible to get in a production experiment—in particular, (i) clear minimal pairs in which the same expression occurs both in focused and in unfocused form; (ii) information about the deviance of sentences in which the ''wrong'' expression is focused; and (iii) fine-grained information about the truth-conditional effects of focus.Our procedure improves in two respects on the standard ''introspective'' method: first, the judgments are repeated and thus their stability over time can be assessed; second, the judgments are quantitative and thus more fine-grained than the standard partition into ''acceptable'' and ''unacceptable'' sentences. Still, our method has important limitations as well. First, because of the small number of consultants, we cannot generalize our claims beyond the native signers we collaborated with. In particular, any differences between our ASL data and our LSF data could be due to robust differences between the two languages or to the sort of cross-signer variation that can be found within one and the same language. Second, we cannot assume that our two consultants used the 7-point scale in the same way; when the same ordinal contrasts between two analogous sentences are found across consultants, this can be informative, but absolute acceptability scores should not be compared directly across consultants. Finally, we could only collect judgments on clusters of prosodic markers that were produced in our target videos, and hence we cannot draw inferences about the acceptability of other clusters that might have been produced under different circumstances or by other signers.ASL data were transcribed by Lamberton himself; LSF data were transcribed by Aristodemo and Santoro, who are not native LSF signers but have experience working with LSF videos, and they consulted with Ducasse in some cases.4 As is standard, sign language glosses are capitalized; IX-i represents a pointing sign toward locus i (usually to realize a pronoun, with IX-1 and IX-2 referring to the speaker and the addressee, respectively); loci affixed to verbs represent agreement markers; a word signed in locus i is glossed as WORDi. For reader-friendliness, we use a revised transcription system (the /-notation) in which sign modulations are indicated by modifying the glosses themselves, while nonmanuals appear above the capitalized glosses, with iconic symbols whenever possible, in the order (1) body changes, (2) head changes, (3) facial expressions (e.g., /}^ ).5Except in if-clauses, nonmanuals were only included on the focused words or in the corresponding controls. Average scores precede each sentence, with full judgments appearing as superscripts between curly brackets in the order in which they were obtained. Hence, 4.2{5 5 3 3 5} refers to an average score of 4.2 obtained from raw scores 5, then 5, 3, 3, and 5. Unless noted otherwise, sentences included in the same numbered example were signed and evaluated as part of the same video and thus form minimal pairs. We have added footnotes when there was more than a 2-point gap among the scores obtained for a given sentence—but even in such cases, the ordinal rankings among sentences with ''well-placed'' vs. ''ill-placed'' focus were very stable.In (3) and (4), we investigated examples in which the letter names B and J count as new in the second clause, while the other letter names count as given. Standard theories (see, e.g., Schwarzschild 1999) predict that focus should be acceptable on B/J but not on the other letters. (3b) and (4b) have focus on B/J, and they are indeed more acceptable than (3a) and (4a), which have focus on the ''wrong'' letter (the (c) and (d) examples are also degraded, as is expected, but they offer less-minimal points of comparison because the signer used slightly different nonmanuals). Importantly, B is a relatively static sign, but J involves movement, namely, a rotation executed by the lower arm.6Since the /-notation is new, we will now explain how it should be read, using (3) as our example.In (3a), A was produced with greater than normal amplitude, which is encoded by boldfacing it (A). It was also produced with a longer than normal hold time, hence _A_ .It was accompanied with a forward body lean, hence the / that appears above A, and it was accompanied with raised eyebrows, hence ^. The same remarks apply to the realization of B in (3b) and of E in (3c): in the latter case, the raised eyebrows were replaced with lowered eyebrows, hence ~ instead of ^. Finally, D in (3d) was realized with greater amplitude and with a longer hold time, but also with speed acceleration, hence the + that precedes it. It was accompanied with a forward body lean, but with neutral eyebrows. The same conventions apply to (4).Turning to the analysis, since the letter names appear in a quoted sequence, we do not expect syntactic movement to affect their realization. Thus, focus is solely realized by prominence, and it is acceptable on new but not on old elements—hence the higher ratings for (3b) and (4b) than for the other sentences, which have focus marking on the ''wrong'' elements. In the acceptable examples, focused B and J involved (i) raised eyebrows; (ii) a forward lean, as in Wilbur and Patschke 1998 and Wilbur 2012; (iii) a longer hold time; and (iv) a greater sign amplitude. As we discuss below, (ii)–(iv) are compatible with a (possibly modified) version of Gussenhoven's ''effort code.'' While (i) also exists in some spoken language focus constructions (Dohen 2005, Dohen and Loevenbruck 2009), its grammatical function in sign language is still debated (see Wilbur 2012 for references, and also Dachkovsky and Sandler 2009 for an analysis of Israeli Sign Language that likens raised eyebrows to H tones in spoken language).In (5) and (6), different (''trembling'') versions of fingerspelled letters are used to realize proper names (thus, we write ANN to transcribe the trembled letter A used to abbreviate this proper name). These names are used rather than mentioned, but since they are semantically conjoined, one also does not expect syntactic movement to be applicable—and the results support the same generalizations as the results found with the mentioned letter names in (3)–(4): focus marking on new elements in (5b) and (6b) is more acceptable than focus marking on old elements, as in the other sentences; and the same means are used to mark focus as in (3)–(4), except that in (5) (vs. (3)) speed acceleration replaces greater sign amplitude.Our generalizations also hold of the mentioned word OR in (7c), which involves a forward lean, raised eyebrows, speed acceleration, and a longer hold time. Here too, focus marking on the new element (namely, OR) in (7c) gives rise to a higher rating than focus marking on old elements (namely, ANN and BILL) in (7b) and (7c) (in addition, the sentence without contrastive focus marking in (7a) is acceptable too).Similar patterns are found in LSF, with some differences: (i) forward lean was not observed; by contrast, (ii) we found regular use of head nods to mark focus.In (9b), the focused element was a proper name PIERRE (related to the noun for 'stone'); it involved a movement with greater amplitude, a longer hold time, and speed acceleration.The same generalizations hold of the phone numbers in (10), where new digits but not old ones can be focused.10In sum, in the examples discussed in this section, contrastive focus was marked in both languages by raised eyebrows, accompanied with forward leans in ASL and with head nods and greater sign amplitude in LSF. Greater sign amplitude was often seen in ASL as well, and longer hold times and speed accelerations were also commonly seen in both languages.We turn to the role of focus in triggering exhaustive readings. In the extant literature, exhaustivity effects in sign language were primarily studied in special constructions—in particular, in question-answer pairs in discourse in Wilbur 2012, and in single-sentence question-answer pairs in Caponigro and Davidson 2011 (a construction somewhat similar to pseudoclefts). In both cases, it is hard to decide whether the exhaustivity effect is due to focus marking per se or to the syntactic/ pragmatic environment it is found in. Here we investigate constructions in which focus marking alone is responsible for the emergence of an exhaustive reading, one that would require the insertion of an exhaustivity operator in the framework of Chierchia, Fox, and Spector 2012.The initial effect is seen in the contrast between neutral and focused OR in (11a–b). Judgments were accompanied with an inferential task to test whether OR was in fact exclusive.In (11a), no exclusive reading is available, which is unsurprising since OR appears in a downward-entailing environment, where embedded implicatures are extremely limited. By contrast, in (11b) the exclusive reading is dominant with focused OR, realized with a forward lean, raised eyebrows,11 greater amplitude, and a longer hold time. These truth-conditional effects are replicated under BET in (12). While BET does not create a downward-entailing environment, for pragmatic reasons it does not invite an embedded implicature; none is obtained without focus, but with focused OR a clear effect emerges.12Interestingly, there are several other ways to form disjunction in ASL. In each case, focus was acceptable, but we obtained inconsistent semantic judgments. Specifically, for a given sentence, there were often inconsistent inferential judgments from one trial to the next under IF, while exclusive readings were obtained under BET; still, we do not conclude anything from the latter fact, as we have more judgments with IF than with BET, and those with IF and BET that were obtained at the same time often gave rise to the same pattern. Such inconsistent results were obtained with OR-WHICH (placed between and before the disjuncts), EITHER-OR (placed between and before the disjuncts), OR-IX-IX (placed before the disjuncts), and OR-V (placed before the disjuncts). We leave this empirical problem for future research.In our LSF data, focused OR triggered an exclusive reading in conditionals (in (13b), and also in (15b) and under BET (in (14));14 here too, LSF differs from ASL in that (i) no forward lean was perceptible in LSF, but (ii) head nods were observed, and (iii) an upward shoulder movement could sometimes be seen.LSF can also realize disjunction with a postposed alternating pointing sign, OR-IXa -IXb. Interestingly, both in a conditional and under BET, the focused version was just deviant, as seen in (15)–(16) (exclusive readings were obtained in the control sentences with a focused version of ''normal'' OR in (15c–d) and (16c–d)).As things stand, we do not know why these effects hold. It is possible, but unlikely, that there is something phonologically wrong with the focused version of OR-IXa -IXb.17 Alternatively, it could be that focusing on OR-IXa-IXb fails to yield a semantic effect and is for this reason prohibited by some condition of economy.18 While we do not know why there should be such a difference between LSF OR and LSF OR-IXa-IXb, we already noted that some ASL disjunctions failed to yield an exhaustive reading; it might be that the same effect is at work in LSF, but that our LSF consultant does not accept idle instances of focus.In sum, in all the examples discussed in this section, exhaustive focus on (standard) OR was marked in both languages by raised eyebrows and greater sign amplitude, accompanied with forward leans in ASL and head nods in LSF, and sometimes with a longer hold time or speed acceleration in ASL. In both languages, a clear truth-conditional effect was seen when standard OR was embedded under if-clauses and under BET, as focus sufficed to trigger an exclusive reading.Our data lead to several positive conclusions about focus in natural language.First, even though many ASL constructions reported in Wilbur 2012 express focus by a combination of syntactic and prosodic means, when movement is inapplicable, sign modulations and nonmanuals can suffice to express focus. The same result was obtained in LSF— which, as mentioned at the outset, is historically related to ASL.19Second, in each language a given cluster of nonsyntactic modifications has contrastive and exhaustive effects alike, depending on the context. Furthermore, our acceptability and inferential tasks showed that putting contrastive focus on the ''wrong'' expressions yielded deviance and that exhaustive focus had fine-grained truth-conditional effects. Plausibly, then, it is no accident that contrastive and exhaustive focus can be expressed by the same means in spoken language—this result appears to hold of language in general, including sign languages.20 It is certainly compatible with the view that focus just signals the activation of alternatives and that further material (contrastive elements present in the discourse, or a covert exhaustivity operator) is responsible for semantic differences between the two cases.Third, the phonological realization of focus in both languages is of theoretical interest:For methodological reasons mentioned at the outset, work with more signers will be needed to establish the robustness of these generalizations.As mentioned, in spoken language higher pitch, greater loudness, and longer duration can be used to express focus. Longer duration is found in sign language as well. One might or might not want to equate greater sign amplitude with greater loudness or speed acceleration with higher pitch. But certainly greater loudness, pitch, and sign amplitude, as well as speed acceleration, can be taken to instantiate Gussenhoven's ''effort code.'' On the other hand, forward leans (in ASL) or head nods (in LSF) are not really explained by the notion of ''greater effort''—after all, one could just as well expend effort by leaning backward or sideways. Similarly, longer duration per se (in both modalities) might not be the most perspicuous way to signal greater effort. A better generalization might be that the spoken and the signed modalities both make use of a ''salience code'' whereby focused elements are made more easily perceptible: holding a sign longer and leaning toward the addressee to realize it are presumably ways to make a sign more easily perceptible.Fourth, in ASL and LSF alike, raised eyebrows are a regular marker of focus—a standard result in the literature on ASL (Wilbur 2012). Future research should (i) compare this finding with the raised eyebrows found as co-speech gestures co-occurring with focus in Dohen 2005 and Dohen and Loevenbruck 2009, and possibly (ii) explore broader cognitive effects of raised eyebrows to determine whether and how they should be integrated into an ''effort'' or ''salience'' code (see Elliott and Jacobs 2013).Several questions remain for future research. First, one should try to replicate our findings with more consultants, to determine whether the differences we found between our ASL and our LSF data are differences between individual signers or between languages;21 similarly, one might use more sophisticated methods to assess the various manual and nonmanual modifications studied here—a nontrivial undertaking (see, e.g., Wilbur and Martínez 2002 on sign speed). Second, one should seek to tease apart the role of different markers, which were lumped together in the present study. Third, the case of exhaustive focus should be revisited to determine why in ASL and LSF alike ''standard'' disjunction gives rise to clear exclusive readings, whereas other varieties of disjunction sometimes display a different behavior. Finally, it would be interesting to determine what the relation is, if any, between the head nods found in our LSF data and the forward leans described for ASL.