Markedness as figure-ground manipulation: a hypothesis

2.1 Figure-ground gestalt (FGG)

FGG refers to a feature of visual perception that entities in a visual field fall under two categories: figure or ground, with the former standing out against the latter. In the study of language, FGG has been invoked to account for a host of linguistic phenomena across languages: the morphology of Greek (Hardy 1992), reflexives in Romania (Cornilescu and Nicolae 2017) and Icelandic (Wood 2014), the dative case in Emai (Schaefer and Egbokhare 2013), masculine noun inflection in Polish (Janda 1997), verbs of removal in Italian (von Heusinger and Schwarze 2006), Grade 2 verbs in Hausa (Abdoulaye 1996), motion verbs in standard Arabic (Alhamdan et al. 2018), applicatives in Tepehua (Watters 1995), topic continuity in Chamorrow (Fox 1982), logical connectors in English (Chen 1991), and the rise of inflectional morphology in general (Dressler 2011).

While these works have demonstrated the utility of FGG, they have not addressed the fundamental question of what FGG is and why it has proven to be so prevalent in language. I contend that the answer to these questions lie in the physiology of the human eye, the neurological limitations of attention, and the relationship between language and cognition.

The most relevant part of the anatomy of the human eye to our discussion is the retina. The retina is the paper-thin tissue that lines the back of the eye and contains the photoreceptors – light-sensing cells of two types: cones and rods – which send visual signals to the brain. The photoreceptors, however, are not mixed or evenly distributed in the retina. On the contrary, the cones, which enable the perception of fine details and color, are concentrated only in the macula, the center part of the retina. Further, there is a pit (or depression) in the center of the macula called “fovea” to provide the highest degree of visual acuity. The rods, which sense low light levels and hence offer less visual acuity – on the other hand – populate the areas outside the macula. So, the entity whose light is “processed” by cones comes into focus to become the figure, and those whose light is processed by rods form the surrounding “ground.” In other words, in a visual field in which multiple entities exist, the physiology of the human eye leaves the perceiver no other alternative than to perceive them as figures or grounds (Chen 2023).

But what determines the figure and the ground in a visual field? Wertheimer’s (1924, 1982) and Koffka’s (1935) respective works led to a set of “gestalt laws” that have been substantiated by later (and a large amount of) experiments. These gestalt laws are “law of similarity”: similar things tend to appear grouped together; “law of good figure”: ambiguous objects are perceived in the simplest way possible; “law of proximity”: points that are connected by straight or curving lines are seen in a way that follow the smoothest path; and others. These principles are then “translated” into a range of linguistic features. A figure tends to be – among other things – “thing-like, solid, discrete, well-defined, tightly organized, small, stable, and symmetric” while a ground – also among other things – is generally “unformed, diffuse, shapeless, less definite, loosely organized, large, unstable, and irregular” (Wallace 1982: 214; see also Eriksen and Hoffman 1972; Reinhart 1984).

Coupled with the physio-anatomical aspect of FGG is the neurological characteristics of attention. If there is one consensus in the massive cognitive psychology literature on attention, it must be the notion of “attentional bottleneck” recognized by Broadbent (1958), limitations in the thinking subject’s capacity to consciously entertain multiple trains of thought simultaneously.^[3] This limitation, to my knowledge, has not been linked to the anatomy of the human eye, and I contend that it should be: that the two are related in some way, although the precise formulation of that connection requires experimental research.

2.2 FGG and language

Although FGG has been applied to the study of a great many linguistic constructions in all areas of linguistics, as was seen above, it is cognitive linguistics that explicitly recognizes the intimate (or rather inevitable) relationship between FGG and language. Among the theories that do so – idealized cognitive models (Lakoff 1987), cognitive semantics (Talmy 2000a, 2000b), and cognitive grammar (Langacker 1987, 1991) – the last is the most systematic and explicit.

First, both Talmy and Langacker recognize the determinative role FGG plays in language. Consider:

Given that these three sentences are identical in syntax – i.e., they are equally grammatical – the unacceptability of Example (3) can only come from the referents of the NPs. That is, the sentence is (nearly) unacceptable because its figure the house is bigger in size than the ground the bike, indicating strongly that the grammaticality of sentences of this kind depends not on syntactical but perceptual and hence conceptual grounds.

Second – and more importantly – FGG is viewed as “a valid and fundamental feature of cognitive functioning” (Langacker 1987: 120). To Langacker, figure is a “substructure perceived as ‘standing out’ from the remainder (the ground) and accorded special prominence as the pivotal entity around which the scene is organized and for which it provides a setting” (120). Due to this “special prominence,” figure shows up as the trajector in the trajector/landmark asymmetry, the source of energy in a billiard ball model (231–236), the causer in a causative construction (cf. Goldberg 1995), or the mover in a motion event (Talmy 2000a). All these roles – further – converge in the codification of figure as the subject of a sentence.^[4]

It should be pointed out that this cognitive view of what the subject of a sentence is is different from the communicative dynamism (CD)/information structure view. On the latter view, the subject is the theme (as opposed to the rheme) of a sentence, one that carries the lowest CD or presents given (as opposed to new) information (Lambrecht 1994). Although the principle of information structure is among the most received theoretical constructs in linguistics, it is ill-suited for the FGG hypothesis due to its treatment of language as a carrier of information, which has two consequences. The first is indeterminacy: the subject presents given information most of the time but not all the time. A tiger in A tiger is a cat can certainly be said to be the theme, but it presents new information. Second, the information status of a linguistic unit depends on context. However, markedness, as is studied here, is anchored in constructions, whose meanings are in general context-independent (Goldberg 1995). In addition, the actual analysis of a construction within the framework of CD/information structure can run into serious problems as seen in, for instance, the study of full-verb inversion (Chen 2003, 2019).

In sum, both cognitive grammar and cognitive semantics view the subject as the sentential designee of figure in the unmarked version of language, which forms a theoretical foundation for the FGG hypothesis advanced in this paper.

2.3 FGG and markedness

If FGG is a determinant of the unmarked structures in the major machinery of language, how can it account for markedness in language? In light of what we know about the marked/unmarked distinction, the unmarked side of language is structured to fulfill typical communicative needs. However, communication will lead to atypical situations. To adjust to these atypical, or “marked,” needs that arise from specifiable contexts, the speaker will manipulate the arrangement of FGG, leading to atypical FGG configurations. Suppose I see a unicorn-like cloud in the sky and want you to enjoy the rare sight. However, it may not be the best for me to say

for I cannot assume that you can identify the unicorn-like cloud among other clouds in the vast expanse of the sky. What I could say, among others, is

I would have just used a full-verb inversion. In Chen (2003), I proposed an FGG-based model of Ground before Figure to account for this construction, proposing that inversion is an instantiation of this model. It enables the hearer to search the ground for the figure so that the figure is ensured the focus of attention when it appears. The Ground-before-Figure model was able to provide answers to a host of questions about inversion that previous accounts, such as information packaging, were not able to explain, e.g., the negation issue (that inversion does not allow negation of the verb, e.g., *In the room is not a unicorn), the transitivity issue (that inversion does not allow transitive verbs, e.g., *In the room chased a unicorn a cat), and the embedding issue (that inversion cannot be embedded in a subordinate clause unless the matrix verb denotes mental activity, e.g., *If in the room is a unicorn, I would not enter).

While Chen (2003) deals with one marked construction, the present paper proposes a hypothesis about markedness in general, formulated as:

FORMULATION 1: FGG Hypothesis on Markedness

A marked construction is a mechanism of FGG manipulation: the rearrangement of the presentation of figure and/or ground in a visual field to fulfill a discourse-specific need that cannot be fulfilled by its unmarked counterpart.

Several notes must be made about the hypothesis proposed here. Firstly, as it has been seen above and will be seen below, the hypothesis is consonant with Langacker’s cognitive grammar and Talmy’s cognitive semantics. However, it is different from them in one crucial way. While cognitive grammar and cognitive semantics recognize the role FGG plays in the unmarked architecture of language, the present hypothesis recognizes the role FGG plays in motivating markedness in language. In other words, the hypothesis is the first attempt to tie FGG, a perceptual theoretical construct, with markedness, a linguistic phenomenon whose centrality in linguistics is uncontroversial.

Secondly, FGG being about the sense of sight, the question arises as to why the sense of sight is relevant to the study of markedness but other senses are not. The answer that might come to our mind first is the privileged position the sense of sight occupies over its cousins. For example, when a verb of sense is used to mean understand, languages invariably select see or its equivalent, not hear, smell, taste, or touch (I can see/*hear/*smell/*taste/*touch how General Relativity can be invoked to define TIME). However, this fact is only the manifestation of the underlying reason. The underlying reason lies in anatomy. As is discussed above in Section 2.1, there are two types of photoreceptors in the retina – cones and rods – that “force” some entities (usually one entity) to be perceived and then conceptualized as the figure and others as the ground. But the organs of other senses do not seem to possess this distinguishing capacity (Drake et al. 2010). Take the sense of hearing. Anatomically, there is no mechanism in the reception and transduction of external auditory stimuli to differentiate sounds in the way visual stimuli are differentiated into figure and ground. Sounds are distinguished only in terms of frequency (pitch) and intensity (loudness), not in terms of salience.^[5]

The third note is on the crosslinguistic applicability of the FGG hypothesis. In the sense that FGG is a perceptual phenomenon, it should transcend the boundaries of language. I believe it does. In Chen (2003, Chapter 5), for instance, FGG is found to account for full-verb inversion not only in English but also in Farci, Chinese, and Chicheŵa (a Bantu language spoken in East Central Africa). In Chen (2009), I argued that the existential construction is yet another figure-ground instantiation, and the universality of the construction is well-recognized in the literature. In this paper, a few non-English languages will be cited also to demonstrate the crosslinguistic application of the hypotheses: Chinese, Italian, Okinawan, and Samoan.

FGG manipulation can take various forms. It can present figure and ground in different orders. It can enfigure an entity that would not be the figure in an unmarked construction. It can select an entity as figure from among a list of candidates. It can also create a figure out of an entity that would not be a figure otherwise. In Sections 3 through 6, I analyze four marked constructions to demonstrate these possibilities.

3.1 Negative-X

In a negative-X SAI, the fronted element is negative, typically (but not always, see below) indicated by the negation marker not. Scholars disagree on whether all types of negation can trigger SAI. On the one hand, many believe that only clausal negation triggers the AS order; local negation does not (Dorgeloh 1997; Erdmann 1988: 71–72; Goldberg 2006: 172; Lakoff 1991: 58; Lakoff and Brugman 1987; Newmeyer 1998: 46). So, Example (8) is believed not to be an SAI because only the noun money is within the scope of negation.

However, local negation can lead to SAI, as attested by Example (9):

The most important feature of negative-X SAIs is the fact that in all the negation-trigged SAIs, the negative particle is not the only thing that is fronted. It is the negative particle together with something else – for nothing in Example (9) – that is moved to the sentence-initial position. As a result, what is fronted is not only negation, but also another element, most often an idea or proposition: not for nothing means for something. The following are two more examples to illustrate the point. The (a) versions are SAIs; the (b) versions are their respective unmarked counterparts.

In each of these, the fronted negative element expresses an idea that seems to be conceptually complete. Not until Kim had finished, for instance, means something happens at a time after Kim had finished. This completeness accords the element the kind of semantic integrity that is needed to be a figure. As a result of fronting and the inversion of the order of the subject and the auxiliary, the status of figure is ensured for these elements because they are unambiguously placed in the position that is prototypically reserved for a figural entity.

What would happen to these figures in an unmarked SVO sentence? The answer is that many of them would not even exist as an entity, let alone being a figure. The figure of Example (10) (a) is not until Kim finished. In the unmarked (b) version, not is the negation marker for the matrix clause while until Kim finished is the adverbial of time. Not until Kim had finished is a combination of the two notions that exist in different places of the unmarked version. It does not exist as a holistic notion. In other words, Negative-X SAIs actually create figures out of the unmarked SVO sentences and then place them sentence initially to ensure their construal as such.

In negative-X SAIs, the negative lexeme may also be adverbials that are negative or near negative in their semantics: never, barely, seldom, and little.

The figures in these sentences are X-ness (the quality of some feature indicated by X): never-ness and barely-ness. As is seen in the (b) versions, these enfigured elements are adverbials of degree. As such, they are abstract in terms of notion and low on the structural hierarchy (not being obligatory elements such as subject, verb, or object but being elements that modify the obligatory verb), hence difficult to stand out as the figures in the unmarked SVO environment. Again, the SAI version turns them into figures by placing them sentence initially.

3.2 Only-X

We now move to only-X, the third type of SAI under discussion.

In Chen (2003), I argue that the uninverted counterparts of these sentences are indeterminate. They could be either (b1) or (b2). Whichever the case, then, and only then in Example (14) and only if individuals cease to be rational in Example (15) cannot be the figure due to their semantically peripheral status (modifying something else as opposed to presenting a central idea). However, once taken out of their respective positions in the unmarked SV sentences and then placed sentence initially in an SAI, they become figures.

3.3 Adverbial-X

There are a few SAI-triggering adverbials in English that seem to defy being grouped into a class.

These tokens have been judged “archaic,” “unnatural and ungraceful,” and “hardly used outside the literary language” (Green 1985: 125, citing others). Chen (2003) proposes that English inversion may have its roots in the Old English verb-second word order rule. Since the X slot of the SAI construction is treated as the figure, we can say that it is the need to enfigure that has kept these earlier constructions “alive.” In other words, if a speaker wants to present many a moonlight night as the figure, they could do that because “history is on their side,” so to speak.

In sum, the position that SAIs enfigure an entity that would not be the figure in the unmarked SV construction receives additional support from the kind of entities that are enfigured. As was discussed above, in Section 2.1, the inherent features of an entity determine the goodness of its figural candidacy. An entity that is small in size and distinctive in shape is a better candidate for figure than one that is diffused and boundaryless. The entities that are enfigured by an SAI are therefore not good candidates to begin with. Not only do they not have shape integrity; they are things that can hardly be said to have shape in the first place, as they are abstract ideas or propositions that are not visible to the human eye. In the visual field in which they occur, therefore, the presentation of them as figures would need special handling. SAI is a mechanism for that special handling, which we call FGG manipulation.

4.1 The FGG account

The FGG account of the passive construction is formulated as having three postulates.

FORMULATION 3: The passive construction

1. presents the patient of a verbal process as the figure,

2. either eliminates the agent or reduces it to being (part of) ground, and

3. denotes a complex atemporal relation

Postulate (a). The passive construction entails a transitive verb, which takes the agent and the patient as its obligatory arguments. In a passive, the patient is placed in the subject position, a position reserved for figure according to Langacker (1991) and Talmy (2000a). In addition and perhaps more importantly, passivization results in the patient being the trajector in the trajector/landmark asymmetry, another hallmark of figurehood (Langacker 1991: 335–343).

Postulate (b). This postulate spells out two possibilities regarding the agent of a passive construction. The first – agent being eliminated – applies to short passives, those that do not include the agent-designating by-phrase. According to Biber et al. (1999: 938, Table 11.9), short passives constitute 86 % of their data in a corpus of 28,000 tokens. There are two reasons for leaving out the agent in a passive: the agent is “unknown, redundant, or irrelevant” (particularly in academic prose) or the identity of the agent is “either not at issue or it does not need to be stated” (938–939).

Also relevant is the fact that there are verbs that do not allow the presentation of agents at all.

This fact goes well with the current account. As was discussed in Chapter 1, entities in a visual field compete for attention. Since human attention is limited, the principle of economy leads to selectivity of entities to be presented in a sentence. The passive construction affords the speaker the opportunity to leave out the agent (for whatever discoursal considerations) so that other entities, particularly the figure, could have more attention. This could be the reason for the high frequency of occurrence of short passives in usage.

The second possibility in Postulate (b) applies to long passives in which both the patient and the agent appear. Consider:

In the active, John is the subject; the door is the object. The sentence is of the SVO structure (Quirk et al. 1985). As such, both John and the door are obligatory (e.g., *John opened). In the passive, however, the agent-denoting John is reduced to the much lower syntactic status of the object of a preposition, similar to other adverbials:

In Example (26) (a), what causes the door to open is the action of the agent; in (b), what causes the door to open is an instrument; and in (d), what causes the door to open is the natural force of air movement. These causes differ in terms – and (possibly) follow a descending order – of human volition and intervention. But that does not matter; they are all coded as oblique: the object of by. In other words, the agent John in Example (25) (b) – enjoys no higher structural status than an action, an instrument, or the act of God. Further, consider:

There are two prepositional phrases in the example: the agent-denoting by John and the manner-denoting with a sledgehammer. There is no clear way to tell which is more important. If anything, the favor is tilted toward the manner adverbial on account of the end-focus principle (Quirk et al. 1985).

Postulate (c). This postulate recognizes the change of construal from the active to a passive construction. In an active voice construction, the construal of the event requires sequential scanning (the conceptualizer scanning the component states in serial fashion). So, Example (25) (a) would invite us to view (“scan”) the verbal process step-by-step – we may imagine John extending his hand, grabbing the doorknob with it, turning the knob, and the door opening. But Example (25) (b) does not. It asks us to scan summarily, so that we can activate the component states of the event “holistically as a single gestalt” (Langacker 1991: 21). In other words, the passive voice turns a dynamic process into an atemporal/stative relation.

4.2 Advantages of the FGG account

The proposed hypothesis has several advantages. Three are discussed below.

Shinzato (2020) discusses an interesting case of “passive warnings” in Okinawan, a Rukyuan language and a sister language to Japanese spoken on the main island of the Ryukyu Archipelago, between Taiwan and the main islands of Japan.

Example (28) is used when the act of hitting is eminent. In such situations, the active voice cannot be used to express the meaning of warning. Postulate (a) in the proposed hypothesis offers a reason. The speech act of warning means that the warned action – “hitting” in this case – will materialize unless the addressee stops the warned behavior. So, the immediate withholding of the warned behavior is paramount to avoid the consequence, making the addressee the natural “focal party” (per Langacker). In other words, the situation demands that the addressee be the figure, hence rendering the active voice – which would present the speaker as the figure – inapt for the task.

Closely related to the passive construction is the middle voice, a construction that is found in many languages (Kemmer 1993).

The challenge posed by the middle voice for a linguistic theory is that its verb is active but its meaning is passive, i.e., the subject is not an agent but a theme.^[7] I argue that the middle voice receives a coherent explanation from the FGG account of the passive on two counts. First, the middle voice is similar to the passive in that it presents a non-agent entity as figure. Obviously, the agent-defocusing account that is popular in the literature (Lyngfelt and Solstad 2006; Shibatani 1989) will not work, as there is no agent in the middle voice to defocus. Second, like the passive, the middle presents an atemporal, stative relation. Example (29) (a), for instance, is not a description of the act of someone selling the model but a description of the fact that the model is a popular commodity on the market. The same may be said about the other two examples: the bread possesses the quality of being cut easily (b) and the book is a smooth read (c).

The third advantage of the proposed FGG account of the passive lies in its formulation that does not tie the passive meaning to a particular structure. We consider two languages – Chinese and Samoan – below. Example (30) is from Chinese; Example (31) is from Samoan.

In a context in which the speaker wants to say that the infant has been fed, the passive voice in (30) (b) – indicated by the passive marker bèi – should be used (Chen and Hu XXXX), as the English gloss suggests. However, the active (a) is the only option for speakers. There are two possible explanations. The first is economy: the fact that infants cannot feed themselves renders the passive necessary. The second has to do with the language being topic prominent (Li and Thompson 1981): the lexeme háizǐ is not the subject but the topic of the sentence. Whichever may turn out to be will not be a problem for the FGG account, as the FGG account allows all kinds of structures to express the passive semantics.

Both sentences in (31) exhibit the canonical VSO order: the subject follows the verb and precedes a direct object. In (a), therefore, e le tama “the boy” and le teine “the girl” are respectively the subject and the direct object of the sentence. In (b), on the other hand, le teine is the passive subject and e le tama is the agent but is neither a subject nor an object. The question for us is: how do we explain the fact that voice is not marked on the verb opo “hug,” as it is the same in both the active (a) version and the passive (b) version? Langacker proposes that this is in line with “a regular pattern that has the semantic effect of deriving passive verbs from active ones but is phonologically null” (Such a proposal suggests the “virtue of the theory [his cognitive grammar account of the passive] that it allows us to describe the phenomenon without being forced to make artificial distinctions” (342). The current proposal, however, offers another – and a better one, I believe – possibility, that the notion of passivity in Samoan is expressed by word order. In other words, while passivity is by and large a morphological notion, it may not be so in all languages.

5.1 The function of it

Jespersen (1937: 89) may be the first person to have noted that it cannot be replaced by any other lexeme. Some of the later scholars acknowledge the fact but none has attempted an explanation. The FGG account offers a natural explanation.

In the visual field represented by an it-cleft, the speaker wants to present as figure an entity that is not (or cannot be) the figure in a default FGG. That entity therefore requires introduction. The ideal part of speech for the task are, naturally, pronouns that point to a referent (instead of denoting a referent directly). Among the pronouns in English, it turns out to be the best candidate. First, unlike most of its peers (e.g., I, we, you, he, she), it may refer to anything and everything in the world. Its plural counterpart, they, could do the same, but plurality limits its (they’s) utility (see below). Second – and perhaps more importantly – it is the only member in the pronominal system that can be both anaphoric and cataphoric.^[10] In a word, it is best suited for the job because it can be a space filler as well as point forward.

The lightness of it in terms of information load is also seen in the fact that there are languages in which no pronoun corresponding to it is used in the equivalent of it-clefts. The following is an example from Italian.

In sum, it serves a double function in the it-cleft visual field. Structurally, it is a space filler for the subject position due to its lack of denotatum. Perceptually (and conceptually), it is a pointer, indicating to the hearer that a figure is about to be named.

5.2 Singularity of BE in the it-clause

Prince (1978a) points out that the linking verb BE in the it-clause is always singular but offers no explanations as to why.

The FGG approach allows for an explanation. Note that the X slot in the it-clause is reserved for the figure. If the figure is five people, five people can be viewed either individually or collectively. As individuals, five people may be different from each other in size, skin color, gender, age. In the visual field in which they are presented as individual figures, they may spread out over a sizable area. However, once they are presented in singular terms, all these differences disappear. They are now one group whose members are presumably thought of as identical to each other. In other words, given the reality that there are four or five persons, Example (36) presents them as a collective, and a collective is a better figure than the individual members inside it. In addition, collectivization of individual figural entities is also found in canonical SV constructions:

5.3 The status of the that-clause

What is the structural status of the that-clause in an it-cleft? The term “relative clause” is used by almost every scholar on the subject, but many of them are apparently uneasy with the categorization, because the that-clause, simply put, is not a relative clause. Consider below, a modified representation of the sentences in Example (33), with the addition of Example (40).

There are two reasons against treating the that-clause as a relative clause. First, if the that-clause were relative, the replacements of that in all these sentences in Example (39) should be acceptable. But most are not, with the clear exception of (a).^[11] Take (b) for instance, if the that-clause were relative, it would be possible to use where instead of that, as where is a designated relative pronoun for a spatial head noun such as London. In fact, in a non-it-cleft context, that cannot be used at all:

The second reason comes from restrictivity. A restrictive relative clause is one that provides essential (e.g., identifying) information for the head noun, which cannot be set off by a comma. An unrestrictive relative clause is the opposite. As such, the that-clauses in Example (39) (a–c) should be unrestrictive, as they provide no identifying information for their respective head nouns (John, London, Ronaldo). However, they cannot be, as they allow no setting off from their head nouns by commas. The strongest case is (a), in which who makes the that-clause every bit look like a relative clause but tolerates no insertion of a comma between John and that/who:

So, what is the status of that? That appears to be a semantically empty connector, much like the way it functions below.

In both, that serves the function of a connector. In Example (43), that connects an object clause with the verb; in (44), that connects a complement with the verb. As a lexeme, that can be said to be a structural marker. By analogy, that in the it-cleft construction can be said to be a conceptual marker of ground, connecting the ground with the figure.

Lastly, if the it-cleft construction is motivated by FGG, and FGG has a physio-anatomical basis, the account offered in this paper should have cross-linguistic relevance. That might be true, as the it-cleft construction or its equivalents are found to exist widely. It exits in, for instance, French (Katz 2000), Chechen (Komen 2015), German (Fiedler 2014), Chinese (Zhang 2014), Czech (Kudrnová 2019), Swedish (Huber 2003), Tagalog (Latrouite 2021), Mayoyao and Ifugao – both spoken in the Philippines – (Hodder 1999), and a few West Chadi languages such as Akan, Ga, and Ngamo (Grubic et al. 2019).