3.1. Some Basics of Cognitive Grammar.

The theoretical framework of CG diverges significantly from autonomist approaches to grammar that assume some form of what Langacker (1991b:515) calls the AUTONOMY THESIS. According to this view, which is not assumed by cognitive linguists, “grammar constitutes a separate, irreducible level of linguistic structure (one with its own constructs, representations, primitives, etc.) that is properly described without essential reference to meaning” (Langacker 1991b:515).

From the autonomist perspective, the semantic characterization of a clause is claimed not to be directly mirrored by the surface morphosyntactic elements, but rather to be affected by interpretive rules of various sorts that mediate between purely syntactic structures and other components (modules) of the grammar, most importantly LF (see chapter 1 in Chomsky 1981 and chapter 9 in Haegeman 1994 for representative treatments).¹²

In contrast to the objectivist truth conditional semantics assumed by many autonomist theories, Langacker (1987a, 1991a, 1991b) proposes what he calls a SYMBOLIC ALTERNATIVE to autonomist approaches residing in the assumption that “grammar is inherently symbolic; only units with both semantic and phonological import are required for its proper characterization” (Langacker 1991b:516). In other words, CG adopts a conceptual view of meaning, whereby semantic structure is equated with conceptual structure-meaning is equated with conceptualization and is assumed to be ultimately reducible to cognitive processing. Consequently, CG assumes that much of grammar is inherently meaningful, not autonomous or accidental, and that “all valid grammatical constructs have some kind of conceptual import” (Langacker 1991a:282).

A fundamentally important feature in CG theory for the elucidation of linguistic structure and meaning in general and for the problem in this paper in particular is IMAGERY, which Langacker defines as the ability speakers have to construe an objective event or situation in different ways for purposes of grammatical coding. Note that imagery is a technical term within CG and is not restricted to the visual realm, but encompasses all aspects of construal. From a practical standpoint it can be taken as asserting that the meaning of a linguistic expression involves not only its objective content, but also how that content is construed. A transparent example to illustrate this is a pair of sentences such as The glass is half empty and The glass is half full, which clearly differ in meaning while describing the same objective situation.

CG posits that grammatical constructions involve the grammaticization (or encoding) of conventional imagery. That is, grammar involves the structuring and symbolization of conceptual content, and a grammar is defined not as an algorithmic device for generating all and only the grammatical sentences of a language, but rather as “a structured inventory of conventional linguistic units” (Langacker 1987a:57). The content requirement of CG is highly restrictive in that the only linguistic units posited in the elucidation of linguistic expressions are the following: semantic structures (predications), the phonological units that symbolize these structures, the symbolic unit itself, structures that are schematic for these units, and categorizing relationships between permitted structures (Langacker 1991b:2). Significantly, no autonomous syntactic level is posited.

Grammatical rules in CG take the form of constructional schemas, that is, schematic “patterns for combining simpler symbolic expressions to form progressively larger ones” (Langacker 1991a:278). Such schemas are assumed to coexist in the grammar along with their instantiations, and are viewed as complex conceptual categories in their own right. The theory does not posit any kind of derivational operations that derive surface structures from underlying deep ones.

There are several aspects of imagery that figure prominently in defining linguistic predications and prove useful in the analysis that follows. First, linguistic predications are defined through the imposition of a figure/ground organization on one or more cognitive domains (complex knowledge structures of varying degrees of complexity). Within a given domain, a particular subpart will be invoked for purposes of linguistic coding. This linguistically relevant subpart, the scope of predication, can be characterized as “the array of content a predication specifically evokes for its characterization” (Langacker 1991a:4).

The PROFILE is the structure within the background (scope of predication) that is singled out for specific designation by a linguistic predication, and the relation between the profile and background (BASE) of a predication determines its semantic value. Profiling is often a matter of degree. Nominals profile things (regions in some domain), and verbs and adjectives profile different kinds of relations among entities. The figure in a relational profile is called the TRAJECTOR (TR). A salient entity in the profile other than the trajector is called a LANDMARK (LM). As an example, consider the two English prepositions above and below. Both prepositions can be used to profile the same objective locative relation between two entities (typically things), but they differ in their TR/LM assignments (and this is a matter of imagery). Whereas the TR of above is construed to be located spatially superior to its LM, the opposite configuration holds for below: its TR is construed as being located under its LM. This resulting difference in imagery results in a difference in meaning that no one would dispute. As demonstrated below, the same kind of theoretical machinery can be used to analyze German impersonals in a way that makes their grammatical structure transparently indicative of their meanings.

Finally, it is important to note that CG assumes that absolute predictabilty is unrealistic and unnecessary in showing that semantico- conceptual structure often shapes and motivates (though does not strongly predict) morphosyntactic structure. “Cognitive grammar does not claim that grammar is predictable from meaning, but rather that it is meaningful because it embodies and symbolizes a particular way of construing conceptual content” (Langacker 1991b:517).

3.2. Image Schemas.

Another important background concept relevant to this paper has been explored in some depth in Lakoff and Johnson 1980, Lakoff 1987, and Johnson 1987, where it is suggested that much of what we call grammar is organized around certain cognitively fundamental pre-linguistic conceptions called IMAGE SCHEMAS, such as container-content, source-path-goal, and link image schemas, to name but a few. Image schemas are assumed to be grounded in everyday physical or bodily experience, and are viewed as mental “structures for organizing our experience and comprehension” that lend “pattern and order to our actions, perceptions, and conceptions […]. [An image schema is] a recurrent pattern, shape and regularity in, or of, these ongoing ordering activities” (Johnson 1987:29).

A fundamental claim of CG is that speakers have the ability to relate and then metaphorically extend prelinguistic image-schematic concep- tions of this sort, which are grounded in a concrete physical domain, to more abstract domains including those relevant to the structuring of conceptual content for purposes of grammatical coding. In other words, image schemas are assumed to “play an important role in structuring cognitive domains […] that support the characterization of basic grammatical constructs” (Langacker 1991b:399).¹³

3.3. Polysemy and the Network (Prototype) Model of Categorization.

Paying careful consideration to the full range of meanings of individual linguistic expressions in human languages, CG views all linguistic expressions, including grammatical structures themselves and gram- matical morphemes such as case, agreement, and expletive elements such as German es, as meaningful and polysemous, with interrelated meanings clustering around prototypical senses, and with senses motivated as semantic extensions from the prototype. Polysemy is therefore recognized as an important aspect of linguistic knowledge that must be represented within the grammar. Each sense of a linguistic structure is viewed as dependent in great part on its particular contextual uses. Some senses may become entrenched, thoroughly mastered structures and thus achieve what Langacker calls unit status (that is, they become cognitive routines, while others may arise as innovations in particular contexts), and may themselves become units if they are activated often enough.

A model of categorization that accommodates this conception of meaning, the network model, is thus employed to represent as precisely as possible the interrelated senses of each morpheme and structure of the language. In such a model, a particular sense of a linguistic predication may be related to others in multiple ways, though usually the set of interrelationships reflects aspects of schematicity and/or semantic ex- tension. That is, one sense may be schematic for another, or it may arise as a semantic extension from another sense when in spite of their perceived similarity its semantic specifications conflict in some way with those of the other (see Langacker 1991a:2–5). The perception of this similarity is then captured by higher order schemas (figure 1). Each semantic network, representing the range of conventional senses of a particular linguistic expression, is conceived as a complex conceptual category, which, as already mentioned, is usually organized around prototypes. It should be noted that the image schemas discussed in the previous section are often schematic for a variety of more specific notions or conceptualizations (subschemas) that may have grammatical import, such as the constructional schemas mentioned earlier (see also Lakoff 1987).

Complex category structure (from Langacker 1993:2).

Note that the categorizing operations involving judgments of schematicity and semantic extension among interrelated senses of a linguistic structure represent two of the restricted array of linguistic units posited within CG mentioned earlier; that is, structures that are schematic for other structures, and categorizing relationships between permitted structures (including those that judge the degree to which one sense or structure within a network is related to another). The network model of categorization has the following advantages: it allows us to represent important information about our conventional knowledge of the lexical items and grammatical structures of a language, which should be included in any analysis that purports to be comprehensive. Further, it accommodates in a natural way the dynamic nature of this knowledge with respect to innovation and change. As I show below, this model of categorization is very useful in elucidating the interrelated meanings and structures of the German es, the DAT case, and the 3sg agreement.

3.4. The Modeling of Prototypical Actions and Events within CG.

Langacker has proposed that our conceptions of prototypical actions and events can be represented in what he refers to as the canonical event model (see figure 2 on p. 96), in which there is “the notion of an event occurring within a setting and a viewer (V) observing it from an external vantage point” (Langacker 1991a:286). A prototypical event is usually thought of as consisting of a chain of asymmetrical interactions initiated by a highly potent, energetic entity (typically an agent) which transmits physical energy toward one or more less potent entities. In our conception of such an action chain, one entity downstream from the agent-induced energy flow (typically a patient) is usually conceived as the energy sink and thus as undergoing some kind of change of state, though there may be other participants in the event (such as experiencers), which are also involved to some degree or another (see chapter 7 in Langacker 1991a and chapter 9 in Langacker 1991b). In figure 2, circles represent entities involved in a conceived event. The transmission of physical energy by the head of the action chain (typically an agent) is represented by a double arrow, and the wavy arrow inside the patient represents the resulting change of state undergone by this entity. The rectangular box encompassing the action chain represents the setting. Profiled entities are conventionally represented in CG diagrams by putting them in boldface (note that in figure 2 the setting is assumed not to be in profile).¹⁴

Canonical event model.

It is of course clear that the canonical event model serves as the conceptual basis for transitive active clauses. However, speakers obviously have the ability to conceptualize other kinds of states and/or interactions involving entities of varying types, so this model is by no means the only one that is operative in organizing how we construe actions and events for purposes of linguistic coding, but in its role as a prototypical conceptualization it can serve as the prototype from which other kinds of less typical conceptualizations can be related.

Let us now consider how a CG account of German impersonal experiencer constructions such as those exemplified in 2–4 above fully integrates the alledgedly meaningless grammatical properties under investigation into a unified meaning-based analysis.

4.1. German “es” as a Pronominal Anaphor (Probable Prototype).

In this section, I discuss the anaphoric function of es . This is perhaps its most fundamental use and the one that clearly has semantic import. I give examples showing some of the various kinds of notions for which it can be anaphoric. The following uses of es are discussed: es as an anaphor for 3sg neuter thing, es as an anaphor for other types of things, and es as an anaphor for reified relational predications.

When es is used as anaphor for a 3sg neuter thing it simply refers to an earlier 3sg neuter entity in the discourse. In other words, it is an action chain participant. The traditional subscript notation shows the relevant reference relationships between the nominal:

In these sentences the pronominal es refers back to an earlier participant in the respective actions, Buch ‘book’ in 5 and Kind ‘child’ in 6, both of which are 3sg neuter nouns.

When used as an anaphor for other types of things, es can replace non-neuter NPs as in 7–8, have a plural NP antecedent as in 9, or be used as a general kind of indeterminate ACC object pronoun (referred to as bloßes formales Objekt ‘mere formal object’ in many German grammars), in which it does not refer to any particular antecedent, but rather to some unspecified thing or concept, as in 10–11.¹⁷

These data show that es is susceptible to semantic bleaching (see Sweetser 1988) even in its basic anaphoric sense: its erstwhile neuter gender specification may be lost, as in 7–8; its singular sense may be lost, as in 9; and in a further extension es may be used to profile no more than some thing, unspecified for gender or number, which has no apparent antecedent in the discourse, as in 10–11. However, in 5–11 es clearly profiles a thing of some kind.

Finally, as a crucial way station toward its more grammaticalized uses es can be used as an anaphor for reified relational predications (and not just action chain participants). In 12–13, es refers back to the atemporal relations expressed by the adjectives arm ‘poor’ and angenehem ‘pleasant’, respectively:

In 14–15, es is anaphoric for reified temporal relations in which the entire processes of reciting in 14 and something having been stolen in 15 are construed as abstract things and referred to by es . Therefore, es is anaphoric for an entire reified action chain conceptualization.¹⁸

Accordingly, the data in section 4.1 demonstrate that es can be used as an anaphor for anything from a thing construed as an action chain participant to a reified clause. Each semantic extension along this meaning chain results in the meaning of anaphoric es becoming increasingly abstract as its antecedent changes from a concrete thing of some kind, as in 5–9, to entities which are construed either as relations, as in 12–13, or reified events, as in 14–15. My claim, which I develop in the next section, is that this latter sense provides the bridge for the use of es to profile settings of various sorts in impersonal constructions.

4.2. German “es”: Profiling Settings of Various Sorts.

In order to motivate the crucial notion that es profiles the setting in an impersonal construction, we need to seriously consider the important fact recognized by native speakers that impersonals have a presentational feel. In other words, they appear to evoke the heightened relevance of the setting or surroundings to the linguistically coded situation. Furthermore, the increased conceptual importance of the setting evoked by impersonal constructions usually coincides with the decreased importance of the participants (if any) involved in the activities construed to take place within the setting. Impersonal constructions thus typically evoke a kind of imagery in which actions or events are described without significantly highlighting or accentuating who or what is doing, or affected by, the action. This is easily seen in weather constructions, such as 1d, in which no agent entity can be mentioned at all, and can also occur in a variant of the agentless impersonal experiencer construction, such as 2–4 above, if the nominal expressing the experiencer participant is omitted, which is often possible:

Impersonal constructions typically provide a way of linguistically encoding an event in which the actor (or agent) participant (if any) has been defocused, though other participants affected in some way by the action (such as experiencers) may be mentioned.¹⁹

It cannot be overemphasized that a characteristic crosslinguistic feature of impersonal constructions is that any notion of an agent participant, if one is relevant at all, is left unspecified (see Langacker, to appear). Consequently some languages provide a means of according another aspect of the scene heightened prominence as grammatical subject in lieu of the missing (or non-salient) agent entity. Recalling the above discussion of the canonical event model, a potential candidate for such special prominence is a more peripheral, though still conceptually salient, non-participant element such as the setting within which the action is conceived to occur.

If we accept the possibility that certain grammatical constructions (for example, impersonals) can accentuate the conceptual importance of the setting within which a state or process is conceived to occur, we must next consider the plausibility of the claim that this kind of meaning can be encoded by a personal pronoun such as es . An immediate concern is the fact that personal pronouns are usually regarded as definite in reference; that is, they typically single out one particular entity in a sentence or discourse. If a pronoun like es is semantically definite, then how can it be used to refer to the setting, which seems to be something that is extremely diffuse and vague in reference?

One key to motivating an analysis in which the pronoun es designates or profiles the setting in German impersonal constructions is that even in their prototypical anaphoric uses the meanings of personal pronouns are inherently schematic with respect to what they refer to: their intrinsic semantic and/or grammatical content basically amounts only to specifications of gender (masculine, feminine, or neuter), number (singular, plural), and case.

Another key to motivating such an analysis is the idea that there is nothing intrinsic in the meaning of a pronoun that necessarily restricts its potential referents to entities of a particular limited size or extent. Just as a lexical item such as place can be used to profile regions in space ranging from the narrowly specific location where a book can be found (The place where I put the book) to the location of a galaxy relative to our own in the universe (The place where the Andromeda galaxy is located relative to the Milky Way), so too can a pronoun’s delimitation- to use Langacker’s (to appear) term-vary with the context in which it is used. Clearly, the larger the region profiled by a noun such as place, the more difficult it is to conceptualize the specifics of the region, but it is still semantically definite. Also, consider how a plural pronoun such as they, which profiles a group of people that excludes both the speaker and hearer in a conversational dyad, has the potential to refer to any group from a few people (They met regularly at the local restaurant) to a group that could plausibly include everyone in the world but the speaker and hearer (They’ll never understand why we chose to give up the reward). Crucially, using a pronoun in a sense that maximizes its delimitation does not necessarily revoke its definiteness. A group designated by they that represents everyone except the speaker and hearer still picks out a particular group. What differentiates such a use of they from one in which the pronoun refers to a smaller group is that its delimitation is large and relatively vague with respect to the actual number of people in the indicated group.

The upshot is that there is nothing intrinsically unreasonable or unmotivated in assuming that es can be used to profile the setting in which an action occurs if we understand that this kind of use represents the endpoint on a scale of increasing vagueness and non-delimitation (see Langacker, to appear). The definiteness of es is still relevant in this use when we take into account that it identifies a unique entity, that is, the setting that is readily discernable in the non-linguistic context even though es has no nominal antecedent. In this regard, Langacker’s remarks about the English impersonal it warrant repeating here, since they are fully applicable to the German es as well:

With this kind of background it should not be surprising that German es could be used to profile the setting, which is maximally diffuse and vague though definite in reference.²¹

The setting es is likely related to the presumed prototype 3sg anaphor and its extensions exemplified in 7–13 above through a series of semantic extensions in which its anaphoric use is gradually broadened, that is, a kind of semantic bleaching to the extent that it loses its association with a participant on the action chain (however indeterminate), and becomes anaphoric for a reified clause or the action itself (see 14–15). From this point it requires only a slight conceptual adjustment to posit a further extension of the meaning of es in which its anaphoric sense is lost completely (that is, it has no apparent antecedent). It becomes associated with the setting in which an action or state is conceived to occur, as opposed to a participant entity in a state or event within the setting, and then actually profiles this setting. It is comparable with the sense of -er in an English word such as kegger, where an erstwhile agentive suffix is associated by a similar process with the event itself. This use of es is easy to regard as meaningless because of its abstractness and its dependence on construal (rather than objective factors), not to mention its recognizable grammatical function as subject. But it is important to note here that from the CG standpoint es has a setting sense and that this is a meaning.

Readers familiar with the work of Lakoff and Johnson (1980) will recognize the sequence of semantic extensions posited above as reminiscent of a kind of metonymy, in which a linguistic predication’s use shifts so that it can be used to designate an entity with which it is frequently associated. In particular, I view the situation with the setting es as a kind of “whole for the part” metonymy (as opposed to the more frequent type, “part for the whole” metonymy, as in We need some new hands around here), where the setting as profiled by es represents the whole within which the part (a state or event) is conceived to occur. Note that such a phenomenon is also known in English. For instance, in a sentence such as The arena witnessed a fierce contest between the fighters, the arena is plausibly construed as a physical setting-subject. Indeed, Langacker (1991b:351 and 1993c) suggests that the setting in which an event is conceived to occur can easily serve as a kind of conceptual reference point with respect to which the event in question can be mentally located.

Es usually structures the scene by setting up a region of some type (often perceived as either physical or temporal) wherein actions, states, or other entities are conceived to be located. In this way the pronominal evokes the container-content image schema discussed in Lakoff and Johnson 1980 and Johnson 1987, with the setting construed as the container and the action or state profiled therein as the content. Such a view is based on the reasonable assumption that humans have the tendency to take a wide-angle view on a scene before focusing upon specific entities or actions within that space.

In the following discussion, I show how the setting senses ofes can be successively linked in a meaning chain from its use as a presentational focusing device or scene setter to a sense in which the imagery evokes the idea that the setting is construed as actually involved in the situation occuring along the action chain. I argue that this is the sense found in 2– 4 above.²²

4.3 German “es”: Profiling Physical or Temporal Settings.

Let us now turn to a sense of es that profiles physical or temporal settings and appears in the clause-initial position only. This instance of es is best viewed functionally as a kind of presentative or wide-angle focusing device that profiles a physical or temporal setting within which some state or action is conceived to take place. Es can evoke, for example, a fairly concrete physical setting, as in the indefinite extraposition constructions in 17a and 18a below (see Perlmutter and Zaenen 1984:171–216). Note that 17a contains the intransitive verb stehen ‘stand’ with a postverbal nominal in the NOM case that is construed as a kind of theme, whereas 18a contains the intransitive verb tanzen ‘dance’ with a postverbal nominal in the NOM case that is clearly an agent. As shown by the ungrammaticality of 17b and 18b, es is unacceptable in any but initial position in this construction type, an issue that is addressed below.

Here the pronominal es can be analyzed as a physical region construed as the locus of a physical relation between two entities, that is, as a setting subject whose function is primarily presentative (analogous to a wide camera angle). Native speaker intuitions support this spatial sense as does the meaning suggested by the typical English translation of such sentences with the deictic there .²³

While es seems to evoke a slightly more abstract physical setting in impersonal passives, as illustrated in 19, its use still seems essentially presentative.

The difference in abstractness between the settings in 17–18 and 19, though probably valid, is actually quite negligible and does not affect the grammatical behavior of the pronoun, since in both clause types it can appear only in initial position. The only difference is that in 17, for example, the region profiled by es is construed to surround a concrete physical contact relation between two specific entities (profiled by the preposition auf ‘on’), whereas in 19 the profiled region seems more diffuse and abstract, specifying merely a vaguely-defined region in which such activities as laughing, singing, or dancing are construed to occur without the necessary overt mention of the participants. The inclusion of the optional prepositional agent phrase in 19c shows that agent participants are conceptually present in this construction, but usually omitted because they are defocused. The slight difference in abstractness of the settings will not have any bearing on the analysis to follow, but does illustrate the polysemous nature of the morpheme.

The examples in 20 show that es can also evoke a temporal setting.

A temporal rather than a spatial setting is most likely involved here because the sentence describes when rather than where an event took place. In this case, the setting is construed not so much as a concrete physical place, but rather as a time span, a kind of abstract temporal space if you will, within which a more specific time adverbial (häufig ‘often’) indicates the frequency with which the swimming activity took place. To bring out more precisely the posited meaning, 20a may be paraphrased as “there was a time within which swimming was often going on.” Similar paraphrases could be proposed for the impersonals in 17–19.²⁴

The scene-setting function of es, as in 17a and 18a, explains why it appears only in the clause initial position (see Smith 1985a). Since the setting profiled by es is mainly presentative, it is not construed as involved in the action per se; that is, it is not construed as being on the action chain. Therefore, its scene-setting function is only logical when it occurs first, whereas mentioning other constituents within the clause first, as in 17b and 18b, eliminates the need for the scene-setting function, since reference has already been made to an entity (or manner adverbial closely associated with the state or action) profiled inside the setting.²⁵

This sense of es is sketched in figure 3, with the light dashed line signifying that the relation between the setting and the action chain consists only in the scene-setting function (see also Smith 1993a:630). The rectangular box representing the setting is in boldface because it is in profile designated by es . Note that figure 3 abstracts away from some of the minor differences between the indefinite extraposition and impersonal passive constructions. The specification of the action chain head in the figure as an agent in the NOM case is only valid for the indefinite extraposition sentence in 18a, since the postverbal nominal in 17a is not an agent. Also, a slight modification would have to be made in the figure for the impersonal passive, since this construction type does not have a NOM-marked agent, but only allows an agent entity to be expressed periphrastically within a prepositional phrase introduced by von, as shown in 19c.

Impersonal scene-setting es construction.

From the CG perspective, in the relational predications profiled in these kinds of impersonal clauses, the setting is plausibly construed as the most prominent entity (that is, the figure in the relational profile) due to its pivotal presentative function as a reference point in shaping how the actions or states are to be accessed by the speaker and/or hearer. We have already noted that in CG the TR is defined as the figure in a relational profile, so the setting in these constructions is analyzed as the clausal TR. Langacker (1991b:305–321) argues persuasively that a schematic definition of the notion “grammatical subject” abstract enough to subsume all of its various instantiations is that of the relational figure of the clause. Thus, the CG analysis concludes that the setting profiled by es in impersonal constructions, that is, the clausal TR, is the grammatical subject in the NOM case. The other prominent entities are LMs.²⁶

4.4. German “es”: Profiling Restricted Settings (Facilitative “es”).

Another kind of setting sense for es is facilitative es . It arises in so-called reflexive impersonal passives such as those in 21, and profiles restricted settings construed as quasi action chain heads.

Despite their name (see Perlmutter and Postal 1984:134-139), these are not analyzed as passives in CG (note that they have no passive morphology). Though the meanings of these sentences evoke the conceptual salience of agentive humans engaging in the designated activities, these entities are completely defocused: the overt expression of an agent is not possible in this construction type. Here the physical setting evoked by es can be motivated as restricted in scope to the area immediately relevant to the action profiled by the clause, and also as being involved in some way with that action as the action chain head, rather than merely as a scene-setter (see Smith 2002). We may therefore speak of such restricted settings as more closely connected conceptually with the other profiled entities than the purely presentative settings illustrated in 17–20 above, and as having a higher degree of perceived potency to affect them. In Smith 2002:89, the term facilitative es is used to describe the pronoun that profiles this kind of setting.

Supporting evidence for this view is the necessity for the manner and place adverbials in 21, which clearly accentuate the importance of the suitability of the setting (place) for the profiled activities to be carried out in a particular manner. The place adverbials accentuate the restriction of the action to a location profiled by the adverbial, and the manner adverbials highlight the suitability of the setting for the profiled activities and its role in facilitating these activities (as though it has quasi-agentive properties). These constructions are judged ungrammatical by native speakers, if the manner or place adverbials are missing. Consider the paraphrase for 21b that brings out the relations posited: ‘The setting enables good working conditions here’.²⁷

The fact that this kind of facilitative es also appears in non-initial position in the clause (see 21b, where the place adverbial hier ‘here’ occupies initial position) reflects this difference in meaning, and so the syntactic difference reflects the semantic one. The claim is that this kind of es is not just presentative, but is construed as involved in some way in facilitating the action profiled by the clause as a kind of quasi action chain head. Its influence upon events is shown in figure 4 by the dark arrow connecting the setting to the action chain (see also Smith 1993a:630). Again, as indicated earlier for figure 3, the setting is in boldface because it is in profile and represented by es . In addition, as discussed in the previous section, this kind of es is also analyzed as the clausal TR due to its conceptual importance in facilitating the profiled actions, so it too is the grammatical subject in the NOM case. Since the conceptual importance of the agent is completely backgrounded (defocused) and thus never expressed, the action chain entity marked as the agent in figure 4 is not specified as taking NOM case.

Impersonal facilitative es construction.

4.5. German “es”: Profiling Settings Construed as Action Chain Heads.

From facilitative es it is only a small conceptual step to a sense of es that profiles settings construed as action chain heads that affect participants on the action chain. The use of es in examples 2–4 above (the impersonal experiencer constructions) may be construed as profiling a setting entity that exerts some kind of physical or psychological effect on an experiencer entity along the action chain, such as mir ‘me’ in 2 (repeated below as 22), as opposed to profiling an entity construed as facilitating an action in general, as in 21 above.

I propose that in this construction type es is also a setting subject, as in 17–21 above, and also the action chain head, as in 21, but that it differs crucially with respect to how its role in the conceived states or actions is construed. In the impersonal experiencer construction, es is viewed as the source of energy directed toward the experiencer par- ticipant in the action, profiled in 22 by mir . Note that this is a matter of imagery and is not necessarily part of the objective situation, but depends on how the situation is construed.

Because motivating this particular variety of es as meaningful in the way just described is central to the thesis of this paper, let us examine in greater depth the above claim; namely, that this es profiles a setting entity construed as acting in some way on the experiencer. Although I am aware that taken by itself and without adequate motivation the claim might seem improbable, I think it is possible to motivate such an analysis with native speaker intuitions and the theoretical assumptions of CG providing the necessary supporting evidence. Three main types of evidence when considered together lead to this kind of solution. I will present two of these immediately below, and take up the third one (which deals with the case-marking of the experiencer) in section 5.

First, we have already seen evidence that es is polysemous, and that its use in the impersonal constructions can be related to its more prototypical pronominal senses. Its purely presentative sense (in impersonal indefinite extraposition constructions and impersonal passives) was motivated in section 4.3 as a plausible semantic extension from its pronominal anaphoric senses, especially the one in which it profiles a reified temporal relation (see discussion in section 4.1). Its facilitative sense as a quasi action chain head was then analyzed as a semantic extension from the presentative sense, since es in the so-called reflexive impersonal passive construction, while still mildly presentative, could also be motivated as more intimately involved in the profiled actions than was presentative es . If we accept this sort of meaning chain linking one sense of impersonal es to another, then impersonal experiencer es likely also profiles a setting of some type.

The next question concerns whether this setting is construed differently from the others we have investigated. This leads to the next piece of evidence: the intuitions of native speakers about the meanings of impersonal experiencer constructions. The native speakers I have consulted about this construction type agree, after careful reflection, that such sentences evoke the sense of an abstract force or influence being exerted upon the experiencer, and that this force or influence in turn affects the experiencer in some way (usually, though not always, negatively). In other words, according to native intuition whatever state or activity is involved in these sentences, such as being cold in 2, being dizzy in 3, or being scared in 4, it is conceived to be imposed upon the experiencer from without and not to originate from within the experiencer. Whatever psychological activity is felt to be located within the experiencer is a reaction to this outside influence, and not the cause of it. If this native intuition is to be taken seriously, then what could conceivably serve as the source of this force but the setting profiled by es ?

Given this conclusion, we have a situation in this clause type in which the setting (es) is construed to manifest behavior ordinarily (and prototypically) reserved for human agents in the canonical event model discussed earlier in section 3.4 and sketched in figure 2. However, while settings are of course neither human nor agents in the strict sense, it is possible for humans to conceive of them in this way, that is, as a kind of abstract energy source. My claim is that the impersonal experiencer construction type represents a semantic extension of the canonical event model that has conventionalized this particular kind of imagery. It makes available for speakers of German a conventionalized structure with unit status that allows speakers at their discretion and depending upon individual circumstances to accentuate the relatively passive nature of the experiencer in these kinds of conceptualizations, while highlighting the enhanced role of external factors (represented by the setting) in inducing the experience. Though it differs markedly from the canonical event model in that the head of the action chain is construed atypically as the setting rather than as a human participant, it nevertheless shares important features with the more prototypical conceptualization which suggest the ultimate relation between the two: both profile activities conceived to take place within a setting in which one entity is construed to exert energy and/or some kind of influence toward another. The difference, of course, lies in which entity is construed as the energy source and the nature of the perceived energy (see Smith 1985a, 1993a, 2002 for additional discussion).

The sketch in figure 5 illustrates the main relationships discussed in the above paragraphs (see also Smith 1993a:631). The setting profiled by es in this construction is also the clausal TR, and thus the grammatical subject in the NOM case. As a matter of fact, in its putative role as the head of the action chain that is conceived to affect a downstream experiencer participant, the subject properties of es are construed as particularly strong in this construction (compared to its role in the other impersonals we have examined). Figure 5 follows the convention of showing the setting-subject in boldface to indicate that it is profiled by the lexical item es . Note that the double arrow now extends from the setting toward the experiencer to indicate the setting’s construal as the source of the energy flow in this construction.²⁸

Impersonal experiencer construction.

4.6. Explaining the Absence of Postverbal “es”.

If es profiles the head of the action chain and is directly involved in affecting the experiencer participant, its absence in examples such as 2b, 3b, and 4b above when another clausal element occupies initial position is unexpected and requires an explanation. By way of contrast, the es in the reflexive impersonal passive construction discussed in section 4.4, also analyzed as a kind of action chain head, does appear postverbally. In other words, the meaning and function of es in the impersonal experiencer construction is clearly more than just presentative, and thus we would predict that, due to its status as action chain head and grammatical subject, es would occur in all syntactic positions usually occupied by subjects, including the postverbal position. How can one explain its inability to appear postverbally?

A partial answer may indeed lie in its presentative function, since the absolute absence of postverbal es in the indefinite experiencer and impersonal passive constructions was explained in section 4.3 as correlated strongly with those constructions’ presentative function, which is only invoked if es appears initially. Impersonal experiencer constructions still have a degree of presentative feel, and it makes sense that initial mention of an entity within the setting (the experiencer) obviates the need to make explicit later reference to the setting itself. However, additional factors are involved. A more careful examination of the data reveals complexities, and sometimes indeterminacies, which deserve consideration. As indicated in note 4, recall for example that for some speakers predicates such as grauen ‘be afraid of’ and schwindeln ‘be dizzy’ allow es optionally in postverbal position in this construction, though it is usually omitted. In what follows, I show how the strong tendency to omit es postverbally in this construction, as well as its occasional optionality (depending upon the clausal predicate), has a functional explanation that can be motivated by semantic factors as a natural consequence of the imagery that it embodies. The key to the explanation lies in the presence of the experiencer nominals in the construction and the meanings of the psychological predicates with which they cooccur.

I propose that the general tendency to omit es in any but initial position in this kind of clause is due to the fact that the psychological predicates structure the scene in such a way that the experiencer participant is construed as an actual or potential locus of perceptual and/or mental energy in its own right, in spite of its non-subject status (Smith 1985a). In other words, the pre-linguistic imagery embodied in this construction type involves a conceptualization with two entities competing for the status of clausal TR, or the grammatical subject, due to their semantic or conceptual factors. On the one hand, the setting is construed as conceptually important due to its conceived influence on the experiencer and the necessity of accentuating the latter’s passive role in the conceptualization. On the other hand, the experiencer itself (almost always human) is construed as potent in some way in its reaction to the force directed toward it from the setting. In spite of this competing motivation, a choice has to be made, and the setting is designated as TR in this conventionalized construction type in order to make explicit the kind of imagery just described. Due to its undeniable active nature, however, there is a tendency for speakers to put the experiencer nominal in initial position in this construction type (as shown in 2b, 3b, and 4b). This tendency may be viewed as a grammatical strategy to accentuate the perceived potency of the experiencer vis-à-vis that of the setting. This is not surprising since humans are more likely to identify with and attribute actions to other humans than to an abstract entity such as the setting. When the experiencer appears in initial position, the canonical position for subjects, its subject-like potential then effectively overshadows that of the setting whose own diminished status is signaled by its not being expressed phonologically (though it can be in some instances with some predicates). However, I would argue that the setting is still semantically relevant (construed as the TR of its clause), and that it is therefore the grammatical subject even when its lexical realization by es is left unexpressed, since the experiencer maintains its DAT case marking and the verb agreement is still 3sg (see also Smith 1993a). These are grammatical facts that only make sense if we maintain that the setting is invariably construed as the clausal TR in impersonal experiencer constructions, regardless of word order differences.

Thus, the omission of es in postverbal position in impersonal experiencer constructions can be explained within CG not only as an aspect of its presentative function, but also as a functional consequence of the natural tension between two subject-like entities when a non- subject, the experiencer, is still deemed significant enough to behave in some ways like subjects. Though the experiencer’s increased semantic significance is not great enough to cause a realignment of grammatical relations (that is, the setting keeps its subject status), there is a syntactic consequence-namely, omission of es postverbally-to mark the difference. The claim is that the syntactic behavior of es once again mirrors its semantic contribution to the clause and can be sensitive to functional factors.

In its recognition of the importance of semantico-conceptual factors in the elucidation of grammatical phenomena, this analysis may suggest an explanation for the fact that some speakers are more likely to allow postverbal es optionally with some predicates but not with others. This property may be due to subtle meaning differences among the psycho- logical predicates occuring in this construction with respect to the strength of their perceived effect upon the experiencer. Though a full treatment of this problem is beyond the scope of this study because it would require an in-depth analysis of the meanings of all the clausal predicates that can be used in this construction, the main idea is relatively straightforward to convey by briefly considering again the examples in 2–4. While postverbal es is generally not possible with the copula plus adjective (kalt ‘cold’) combination in 2, it is possible for some speakers with grauen ‘be afraid’ in 3 and schwindeln ‘be dizzy’ in 4.²⁹

Even a cursory consideration of the meanings of these predicates indicates that the verbs grauen ‘be afraid’ and schwindeln ‘be dizzy’ imply a somewhat more direct and forceful effect upon the experiencer by the setting subject es than does the copula plus adjective sequence profiling the state involving kalt ‘cold’. The obvious difference in grammatical coding in this instance, that is, the verb in 3–4 versus the copula plus adjective in 2, reinforces this observation in that the verbs in 3–4 profile actions not states.³⁰

Therefore, the tendency of es in impersonal experiencer con- structions not to appear postverbally would seem to be semantically motivated. But it should also be emphasized that despite the fact that the optional occurrence of postverbal es with some predicates seems to correlate with the meaning of that predicate, it is not possible to strongly predict which predicates will allow optional postverbal es, nor even to predict on a given occasion whether or not es will appear with those predicates for which it is allowed. This is due to the fact that the previously discussed tension between the setting and experiencer in this construction differs depending upon the predicate involved. The conceived relative importance of the setting and experiencer in each instance is not absolute, but a matter of degree. Each variant, therefore, likely reflects a subtle difference in the degree of profiling of the setting vis-à-vis the experiencer. Though the inherent indeterminacy, and thus unpredictability, of this phenomenon is troublesome to those who seek clear-cut rules for all linguistic phenomena, it is a fact about the language that cannot be avoided, and a consequence of imagery that has to be taken into account.

From the CG point of view, the syntactic variants of the impersonal experiencer construction are thus semantically distinct, with each assumed to have its own constructional subschema to sanction particular instances of that variant. The theory posits, for example, a separate subschema under a more abstract impersonal experiencer schema for each of these variants of 4: Es graut mir, Mir graut, and Mir graut es (or possibly Mir graut’s) for those speakers who tolerate postverbal es in this construction. Their clear relation to each other is evident, since each evokes basically the same conceptualization of a setting profiled by es construed as an abstract force or influence that affects an experiencer in a particular way (causing fear). The main difference among the variants lies in how strong that effect is construed to be in comparison to the experiencer’s reaction, represented syntactically, as we have seen, by differences in linear order and the presence or absence of es . Any overt mention of es implies a slightly greater conceptual salience of the setting and a concomitant lesser salience of the experiencer’s role; omission of es implies just the opposite-the setting is downplayed and the experiencer’s role is highlighted.³²

The relationship among the variants comprising the conceptual category represented by the impersonal experiencer construction can be captured by positing a superschema subsuming each structural variant which abstracts away from their syntactic differences and represents those features that they share: a profiled setting construed as head of the action chain, a psychological predicate cooccurring with a DAT-marked experiencer, and 3sg agreement. A rough picture of this category is given in figure 6 (see page 122), where I attempt to represent the relative differences in the degree of profiling of the setting versus the experiencer by differences in the darkness of the boldface used to represent these entities. The strong salience of the setting when es is in initial position is therefore represented by a very dark boldface in contrast to that used for the experiencer (Es graut mir), whereas just the opposite is shown when es is missing and the experiencer is highlighted by being clause-initial (Mir graut). The intermediate case (Mir graut’s) is represented in the middle diagram in which the relative salience of both entities is more equally distributed, as reflected in the fact that es (sometimes in a phonologically reduced form) is possible for some speakers. Note, also, that this kind of representation elegantly accounts for the fact that some speakers do not tolerate postverbal es in this construction: they simply do not have the intermediate subschema in their grammar.

Impersonal experiencer construction category.

4.7. A Preliminary Network Model of the German “es” Category.

Before I turn to the evidence provided by the experiencer’s DAT case marking for analyzing es in impersonal experiencer constructions as a setting construed to affect the experiencer in some way, I briefly consider a rough picture of what the es category looks like up to this point. In figure 7, I sketch a rough approximation of the structure of this conceptual category, as motivated to a great extent in this paper; namely, in its anaphoric senses and the setting-subject senses. These senses have been enclosed within a double lined box. However, it should be noted that the picture of the es category as motivated in this paper is by no means complete. As shown in figure 7, other uses of es are cataphoric. They evoke various kinds of more abstract settings similar to Fauconnier’s (1985) mental spaces which reflect a shift to other kinds of abstract cognitive domains, including those using the pronominal as a kind of iconic distancing device in the grammar for separating one part of a clause from another. Although the cataphoric senses are not relevant to the current study, I include them in figure 7 (see page 124) in order to give a rough idea of the structural complexity of the entire es category. Solid arrows represent relations of schematicity, and dashed arrows represent semantic extensions from one sense to another.³³

We are now in a position to demonstrate how the assumption thates profiles the setting allows us to explain and motivate the DAT case marking on the experiencer nominal in the impersonal experiencer construction.

Proposed es category.

5.1. DAT Objects in Ditransitive (Three-Place) Constructions.

This kind of DAT marks the nominal as a sort of experiencer construed in particular as a possessor and/or recipient of another entity:

Note that the possessor/recipient in 23, Kind ‘child’, is affected by the movement of the book into its sphere of influence, but also that it exerts (or comes to exert) control over that entity. Each of these distinct aspects of the child’s role in the event is clearly significant to the meaning of the clause, and the DAT reflects this as a marker of a particular manifestation of bilateral involvement.

5.2. DAT-Marked Nominals in Various Two-Place Constructions.

In this section, I present data from several two-place constructions that contain DAT-marked nominals manifesting experiencer properties evocative of bilateral involvement, and also NOM-marked nominals functioning as grammatical subject. While space limitations allow neither a detailed listing of examples, nor a fine-grained treatment of how the imagery of each type of construction differs from the others, the following examples should suffice for showing that the DAT in each instance evokes some aspect of bilateral involvement (see Smith 1993a for more details).

It has been argued elsewhere (Smith 1987, 1993a, 1993b) that the DAT objects required by a subset of German verbs (DAT verbs) are not arbitrary, but can be semantically motivated. A careful consideration of the clausal semantics shows that they have properties characteristic of experiencers and thus manifest bilateral involvement in some way. One kind of DAT object is conceived as an actual or potential secondary actor in its own right in response to the action instigated by the clausal TR (subject), as shown in the following example with the DAT verb helfen ‘help’.

This kind of object can be said to have its own initiative potential, given that the teacher’s help allows the child to do something on its own, that is, the help is conceived to induce a reaction in the child. A similar analysis could be given for other DAT verbs, such as folgen ‘follow’, in which the object’s potency may be reflected either in its ability to move as a result of the subject’s pursuit, as in Der Polizist folgte dem Dieb ‘The policeman followed the thief’, or in its perceived superiority to other alternatives, as in Wir folgten seinem Rat ‘We followed his advice’. Note that though this secondary actorhood is not specifically profiled, it is still within the scope of predication of the DAT, and thus part of the DAT’s meaning.³⁷

Another kind of DAT object is illustrated in 25, where the verb trauen ‘trust’ profiles a particular kind of asymmetrical relation between two persons in which the object is not construed as totally passive within the profiled relation, but as manifesting properties often characteristic of agent-like participants, though it is not the grammatical subject of the clause. That is, in being regarded as trustworthy, for example, the child is construed to possess knowledge or have an active potential that the subject can rely on.

The bilateral involvement evoked here therefore involves the sense that the DAT-marked object of the verb is acted upon by being the object of the subject’s trust (its passive aspect), but it is also construed as having properties worthy of that trust (its active potential). Other DAT verbs evoking similar kinds of imagery include glauben ‘believe’, gehorchen ‘obey’, dienen ‘serve’, etc.³⁸

The construction illustrated by 26a,b evokes a somewhat different kind of imagery in that from a purely conceptual viewpoint the DAT- marked experiencer would be expected to appear as the grammatical subject in the NOM case, given that it is the only human entity in the profiled relations and clearly the primary source of energy. However, it is the other nominal, a kind of theme, that is coded (non-canonically) as subject in the NOM case:

This construction type, sometimes called the DAT subject construction to indicate the subject-like potential of the experiencer in spite of its case marking, represents a kind of marked coding in which the expected array of grammatical relations and case assignment is bypassed in order to accentuate a construal in which the perception of another entity by the experiencer induces some kind of evaluative reaction, as in 26a, or other perceptual reaction within the experiencer, as in 26b (see Smith 1993a for an extensive discussion from the CG perspective). In 26b, for example, the experiencer’s perception of the melting ice cream (its active role) plausibly results in a reaction or realization of the change in the ice cream (its passive role), and the DAT case explicitly codes the relevance of both of these aspects of the experiencer’s bilateral involvement in the event.³⁹

5.3. DAT Experiencers in Setting-Subject Constructions.

Given the analysis I have proposed, it should now be apparent that the impersonal experiencer construction evokes a kind of imagery in which the experiencer is bilaterally involved. Recalling the discussion in section 4.5 above and summarized in figure 5, native speaker intuition about the meanings of these constructions confirmed the conceptualization of an abstract/diffuse entity of some sort (that is, the surrounding environment or setting) exerting a force or influence on the experiencer, which in turn induces a psychological and/or physical reaction in the experiencer. We have argued for an analysis in which the pronominal es profiles a setting subject, construed as head of the action chain, that is conceived to affect the experiencer in just this way. Therefore, it is clear that the DAT case marking on the experiencer participant in this construction evokes the same general kind of meaning (bilateral involvement) evoked by the other uses of clausal DAT discussed earlier. The experiencer in this construction is a LM entity construed to lie downstream from the clausal TR (the setting). Apart from the fact that the TR is construed as the setting, the experiencer’s role is very similar to that of other experiencers we have discussed. The DAT is certainly not quirky or unmotivated when its sense here (to mark experiencers of an impersonally-induced action) is related to its other senses in the German DAT category. Its occurrence is rather an integral part of the meaning of the construction, which itself is a conventional unit in the grammar, that is, a thoroughly entrenched and mastered structure.

5.4. The Meaningfulness of DAT and the Setting Subject Analysis.

The above discussion shows that the various occurrences of DAT case used to mark experiencers are not semantically arbitrary, but meaningful in evoking aspects of the nominal’s bilateral involvement in the action profiled by the clause. Each kind of bilateral involvement differs slightly from one construction to another, and a fine-grained analysis would even show that individual senses of DAT within a single construction type probably differ, depending upon nuances of meaning of individual clausal predicates. The individual senses of DAT examined here, including its sense in the impersonal experiencer construction, form a subfamily of senses related to each other in a semantic network (via paths of semantic extension and/or schematicity) which instantiates the bilateral involvement schema. This schema itself is then only part of a much larger and more complex DAT conceptual category that includes other clausal and prepositional senses that can be represented by a still larger semantic network.

Analyzing the DAT as evoking some aspect of bilateral involvement therefore reinforces the setting-subject analysis presented earlier in this paper, and provides the third main argument for this analysis. As a matter of fact, it is clear that from the perspective of CG the occurrence of DAT marking on experiencers and the so-called expletive es in the impersonal experiencer construction are mutually supporting from a semantic point of view, given the assumption that both are meaningful.