A developmental study of prepositional phrases in Hebrew written text construction

Structure and syntactic roles

Traditionally, PPs have been viewed as a kind of case marker, or realizations of features on NPs (Fillmore, 1968; Schachter & Shopen, 2007). The current study is based on a different view, where prepositions are considered heads of syntactic phrases (Jackendoff, 1973), and PPs are treated as unified functional wholes, with their own structural characteristics, and their own unique set of roles (Halliday & Matthiessen, 2004).

Structurally, a PP consists of the head preposition, followed by a complement. English as well as Hebrew prepositions are most typically complemented by NPs, but other complements are found as well, including adverbs and adjectives (Huddleston & Pullum, 2002; Ravid & Shlesinger, 2000), e.g. until recently; me-xadash ‘from-new.SG.MASC = anew’. ¹ Testifying to the intimate relationship between prepositions and their complements is the fact that PPs are prone to undergo fusion and grammaticalization processes, whereby they transform from compositional expressions to semantically bleached, opaque closed-class items (Nir & Berman, 2010), e.g. le-yad ‘to-hand = beside’.

Hebrew prepositions are structurally unique in three major respects. First, the four basic Hebrew prepositions – be, ke, le, and me (roughly corresponding to English in, as, to, from respectively) – are bound ‘wordlettes’, that attach to the word following them as prefixes, often with morpho-phonological consequences (Ravid, 2012), e.g. be ‘in’ + kavana ‘intention’ = be-xavana ‘intentionally’. Second, in cases where these prepositions (with the exception of me) are followed by a definite noun or adjective, they incorporate the definite article, which is indicated by a vowel change, e.g. le ‘to’ + ha-yam ‘the-sea’ = la-yam ‘to-sea.DEF’. Finally, all Hebrew prepositions, when followed by a pronoun, incorporate that pronoun as a kind of inflectional suffix (Berman, 1978), e.g. be ‘in’ + ani ‘I’ = bi ‘in me’.

Functionally, PPs are extremely versatile constructions. The role most typically associated with PPs is the complementation of verbs as ‘indirect objects’ (Givon, 1993) (cf. ‘arguments’ below). But as was demonstrated above, PPs do much more than just that. Some PPs function as modifiers of various parts of speech, including verbs, nouns, adjectives, and prepositions (Huddleston & Pullum, 2002). Others serve as predicates (Hengeveld, 1992). Still other PPs constitute discourse markers (Maschler, 2009). Examples of the various PP roles are given in Table 1. The current study aimed at mapping the different PP roles in Hebrew, with a developmental, discourse-oriented perspective.

OPEN IN VIEWER

Table 1.

Syntactic roles of PPs.

Role	Example
Argument	ben^kita sheli paga  bi member^class of.1SG hurt.3SG.MASC in.1SG ‘A classmate of mine hurt me’
Verb-adjunct	hayínu tsrixim lisbol  biglalam was.1PL need.PL.MASC suffer.INF because.3PL.MASC ‘we had to suffer because of them’
Noun-modifier	ha-métax  beyn xaverot tovot the-tension between friends.FEM good.PL.FEM ‘The tension between good friends’
Predicate	hu lo  ba-rama shelánu he not in-level of.1PL ‘He is not in our level’
Discourse marker	le-mashal to-allegory ‘for instance’
Adjective-modifier	de’ot politiyot hafuxot  mi-sheli views political.PL.FEM oposite.PL.FEM from-of.1SG ‘political views opposite of mine’
PP-modifier	be-érex  ba-émtsa in-value in-middle.DEF ‘roughly in the middle’

Discourse-embedded functions

PPs are frequently mentioned as a means for achieving a dense and concise style of expression, by allowing speakers – and more often writers – to pack several proposition into a single clause (Beers & Nagy, 2009). This compressing ability seems to be central to the functioning of PPs within discourse. But what is it about PPs that allows them to do that? Halliday and Matthiessen (2004) explain the compressing function of PPs in terms of a grammatical metaphor – that is, a realignment in the mapping pattern of semantic units onto grammatical ones. Specifically, a semantic conflation of two situations (events, states, activities), prototypically expressed by a sequence of clauses, can alternatively be expressed – through grammatical metaphor – with a PP operating in the bounds of a single clause. For instance:

(2) a. My client went ahead with the lawsuit even though I advised him otherwise.

  b. My client went ahead with the lawsuit against my advice .

(3) a. Take a large spoon and scoop the avocados from their shell.

  b. scoop the avocados from their shell with a large spoon .

In (2b, 3b), one situation is downgraded, and construed metaphorically as a component part of the other. The choice between clause-combining and clause-internal prepositional modification has to do with the ‘weight’ that is to be assigned to the alluded-to situation, in terms of its referential content, and its communicative and textual status; a PP, unlike a clause, cannot itself be circumstantially elaborated, and does not constitute a speech act, or carry thematic structure of its own. Instead, it offers informational compression, as well as a tighter integration of situations, and an enhancement of the asymmetry in their respective weights.

Of course, compression of two clauses into one, when looked at from the opposite direction, is actually an expansion of that one clause; PPs often enrich the content of a clause by offering elaboration beyond its basic semantic core – in the form of additional participants and/or surrounding circumstances – while naturally also making it longer and structurally more complex.

In their role as noun-modifiers, PPs are crucial for another common kind of grammatical metaphor, namely nominalization. Here the metaphor involves construing situations as static objects rather than dynamic processes (Halliday, 1993). This – perhaps more than any other feature – is considered a hallmark of written and academic language (Biber & Gray, 2010; Nagy & Townsend, 2012). The flexible nature of PPs allows them to be inherited from verbs down to their corresponding deverbal nouns, thus preserving information about participant entities and surrounding circumstances of situations, even when they are realized nominally, e.g. his intrusion into my conversation .

Later-language development

The study of language development has traditionally focused on early acquisition stages, during the first few years of life (Ravid & Zilberbuch, 2003). Research done over the past two decades, has broken beyond this age-span, revealing language development to be a protracted process, extending throughout the school years and into adulthood (Berman, 2004a; Nippold, 2007). Concomitantly, the framework of language development research was extended with respect to modality, to include not only spoken, but written language as well, acknowledging that coming to master the unique discourse style of writing is a major aspect of later-language development (Ravid & Tolchinsky, 2002). With respect to PPs in particular, developmental research has focused mainly on the initial emergence of prepositions in toddlers (e.g. Johnston, 1988; Rice, 2003; Tomasello, 1987), leaving later developmental stages generally unexplored. The current study aims at tracing the developmental course of PPs beyond initial emergence and on to mature mastery (Berman, 2004b), focusing on written discourse. It was hypothesized that PPs will become more prevalent with age, reflecting a growth in clause-internal structural and semantic complexity.

Form–function relations

One of the major insights from research into later-language development is that advanced linguistic abilities do not necessarily entail the acquisition of more and more new forms, but rather the establishment of new form–function relations, as already existing forms are recruited for designating new functions (Berman, 2007). PPs typically appear in children’s utterances during the third year of life (Dromi, 1979; Johnston, 1988), and they hence undoubtedly constitute ‘old forms’ by the time school children start producing written discourse. The developmental challenge that remains is not in using the form ‘PP’ in itself, but in flexibly deploying this form for its full range of different potential functions within discourse, e.g. for noun and adjective modification as well as for verbal complementation. Accordingly, it was hypothesized that the set of PP roles will become more diversified with age.

Usage-based approach

Rule-based approaches to language view grammar as the set of rules which govern the construction of grammatical sentences, and language acquisition as the rapid process of learning these rules (Berwick, Pietroski, Yankama, & Chomsky, 2011; Wexler, 1982). In contrast, usage-based approaches perceive grammar as a set of resources for achieving communicative goals within social contexts, and language development as the lifelong process whereby generalized representations emerge out of individual usage-events of language production and comprehension (Bybee, 2006; Lieven, 2016; Tomasello, 2003). Usage-based approaches naturally give rise to empirical research designs, which consider language use within real-life discourse, rather than relying on decontextualized, structured experiments or constructed examples (Chafe, 1994; Hickmann, 2003; Ravid & Berman, 2010). Adopting a usage-based approach, the current study aims to examine how Hebrew speaker-writers’ use of PPs in the service of extended discourse production is affected by age and the accumulation of linguistic experience, as well as by different contexts of discourse production.

Two central factors in determining the communicative context within which discourse is produced are modality and genre, and these are predicted to affect the text’s syntactic features including the use of PPs. Regarding modality, concern here is with writing, a mode of communication marked by the detachment between addresser and addressee (Olson, 1994). Writing prevents the use of paralinguistic and non-linguistic means of expression, forcing writers to rely on purely linguistic devices, while at the same time enabling the recruitment of greater linguistic complexity, by not posing any immediate-processing demands (Ravid & Tolchinsky, 2002). PPs were found to be more prevalent in writing versus speech (Chafe & Danielwicz, 1987).

Regarding genre, the current study focuses on narrative and expository discourse. While narratives concern agents in a chronologically-unfolding sequence of events (Berman & Slobin, 1994), expository essays present a logically-arranged network of abstract concepts and arguments (Britton, 1994). With respect to the topic at hand, narratives were expected to have a relatively high prevalence of verb-modifying circumstantial PPs, for qualifying surrounding aspects of the narrative events, whereas expository discourse was expected to induce a frequent use of noun-modifying PPs, accompanying nominalizations, and qualifying complex concepts realized in heavy NPs.

PP prevalence

All PPs were identified and counted. Each preposition, along with the phrase or phrases forming its complement, was considered one PP. PPs found inside another PP were also counted, e.g.

(4) sixsux [ beyn ha-banot  [ ba-kita ]]

conflict [between the-girls [in-class.DEF]]

 ‘ A conflict [between the girls [in the class]]’

For identifying prepositions, and distinguishing them from other word classes (Schachter & Shopen, 2007), general and Hebrew-specific distributional, functional, and morphological criteria were set: distributionally, prepositions can take NPs as their complements, and verbs as their ‘external arguments’ (Rauh, 2002); functionally, prepositions head phrases which can serve as verb complements and/or adjuncts; and morphologically, Hebrew prepositions inflect for number, person, and gender (Shlesinger, 2000). Applying these criteria lead to the identification of 45 different Hebrew prepositions in the study texts.

Naturally, this set of Hebrew prepositions does not fully coincide with corresponding categories in other languages. Thus, some function words considered prepositions in English were not considered as such in the present analysis – and vice versa, based on the Hebrew-specific criteria presented above. An example of exclusion is the function word shel, equivalent to English of in such expressions as types of people; this was precluded from the present analysis since it never complements or modifies verbs, and serves only to modify nouns (Berman, 1978; Ravid & Assulin Tzabar, in press) – thus failing to satisfy our distributional and functional criteria. An example of inclusion is the word biglal, corresponding in meaning to English because. While because does not function as a preposition in English, but as an adverbial subordinator, biglal satisfies all three criteria, and thus was considered a preposition (as in other accounts, e.g. Glinert, 1989), and consequently included in the analysis (see second example in Table 1).

The Hebrew accusative case marker et was the only item precluded from analysis despite satisfying our criteria. Accusative et is different from all of the items which were identified as prepositions in that it obligatorily precedes any definite direct object, while failing to precede non-definite direct objects (Berman, 1978).

Syntactic roles

Each PP was classified in one of seven syntactic roles. The different roles are presented in Table 1, with examples from the study texts. The first two roles – argument and adjunct – are both verb-related. The distinction between the two is notoriously problematic: rather than being dichotomous categories, they define the two ends of a continuum, along which numerous kinds of intermediate entities are found (Givon, 1993; Hopper & Thompson, 1980). A more fine-grained categorization of verb-related PPs can be found in Brandes (2015, in Hebrew). For the purposes of this report, however, all verb-related PPs were classified as either arguments or adjuncts. Assignment to these categories was based on semantic, rather than grammatical considerations (Fillmore, 1994); arguments were defined as PPs which designate participants – that is, entities and concepts taking part in the scenario defined by the verb – regardless of the extent they are obligatorily ‘licensed’ by that verb. Adjuncts, on the other hand, designate surrounding circumstances of scenarios, relating to such aspects as time, place, cause, and manner.

Discourse markers are non-referential elements, which rather than referring to things in the ‘real world’, relate to the textual or interpersonal structure of the discourse itself (Maschler, 2009). PP-modifiers are specifiers like English straight in straight through the wall, which in Hebrew typically take the form of fused, semi-grammaticalized PPs.

Verb-adjuncts

Adjuncts are verb-modifying PPs that introduce additional circumstantial information into a clause, thus enriching its content, and making it longer and structurally more elaborate. As such, they constitute important contributors to the kind of clause-level complexity characteristic of mature expression. Concomitantly, adjuncts promote overall informational compression and density, by packing entire propositions as component parts of other clauses.

Unsurprisingly, adjuncts became more prevalent with age, being used twice as frequently by adults and high-schoolers compared with junior-high and grade-school students. The mean number of adjuncts per clause increased from just 0.18 in grade 4 to 0.43 in adults – from less than one in every five clauses to more than one in every 2.5.

The proportion of adjuncts out of all PP tokens increased from about 20% in grade 4 to roughly 30% in adults – nearly one in every three PPs. However, this difference was not statistically significant. The reason is that alongside adjuncts, there was an increase in the prevalence of other types of PPs, so that their relative place in the overall picture was retained.

Recall that adjuncts were hypothesized to be more prevalent in narratives than in expositories, since the kind of concrete events comprising the plot of a narrative were thought to be more readily enhanced circumstantially compared with the abstract concepts dealt with in expositories. However, the two genres did not differ in the prevalence of adjuncts, suggesting that these PPs serve equally important purposes in expository discourse as they do in narratives. The proportion of adjuncts was indeed greater in narratives, likely due to a greater prevalence in expositories of other types of PPs, such as noun-modifiers and discourse markers.

The adjuncts found in the sample were of four main semantic types – place, time, cause, and manner – of which the first two were by far the most common, in both genres. In narratives they indicated specific times and locations, serving three main functions; some designated spatial and temporal aspects of specific events in the plot-line, e.g.

(5) hem tsa’aku et ze  ba-kita

they.MASC shouted.3PL ACC this in-class.DEF

‘The shouted it in the classroom’

Quite a few others appeared within the opening segment, establishing the settings for the entire story, e.g.

(6)  be-sof kita tet  asíti mesiba  etsli ba-báyit

in-end grade nine did.1SG party at.1SG in-house.DEF

‘At the end of ninth grade I threw a party at my house’

A third common function of (preposed) time adjuncts was to mark the start of a new episode, locating it along the narrative’s temporal axis, e.g.

(7)  axarey kama dakot  báti lehikanes la-xéder

after few minutes came.1SG enter.INF to-room.DEF

‘After a few minutes I went to go into the room’

Time and place adjuncts were commonly used in expositories as well. In accordance with the nature of the genre, they were not specific, but rather expressed generic times and places, typically indicating the scope of abstract phenomena, e.g.

(8) lo neyaxasim le-xax maspik xashivut  ba-xevra

not ascribe.PL.MASC to-thus enough importance in-society.DEF

ha-yisra’elit

the-Israeli

‘It is not ascribed enough importance in Israeli society’

While time and place adjuncts were common in all age groups, some developmental trends were apparent, relating to the emergence of a wider variety of complex temporal relations, e.g.

(9) le-órex kol ha-bikur

to-length all the-visit

‘throughout the visit’

of abstract locations, e.g.

(10) ba-xevra

in-society.DEF

‘in society’

and of high-register expressions, e.g.

(11) im xazarato

with return.3SG.MASC

‘upon his return’

A discussion of the other semantic types of adjuncts is beyond the scope of this article. For a detailed discussion of manner adjuncts, see Brandes and Ravid (2016).

Noun-modifiers

Learning to construct and use heavy NPs to provide rich descriptions of entities and abstract concepts is a major aspect of later syntactic development, and especially in Hebrew (Ravid & Berman, 2010). Previous studies have traced the development of various noun-modifying devices such as adjectives, relative clauses, and N-N compounds (Nippold et al., 2005; Ravid & Levie, 2010; Ravid & Zilberbuch, 2003), to which this study adds the PP, highlighting its role as another important strategy for NP expansion.

In grade 4 texts, noun-modifying PPs were still uncommon, appearing only once in every 10 clauses on average. However, their prevalence nearly quadrupled during the age-span investigated, reaching in adults a mean of 0.38 per clause – one noun-modifier in every 2.6 clauses. The proportion of noun-modifiers also increased, from just 12% in grade 4 to a quarter of all PPs in adults – showing that noun modification gradually became one of the major uses for PPs. These results suggest that PPs play an important part in the process of developing NP complexity in written Hebrew.

Using PPs as noun-modifiers is a marked option, since PPs are most basically verb- related, while nouns are typically modified by adjectives. Thus, an increased prevalence of noun-modifying PPs reveals increased linguistic flexibility, expressing the kind of realignment of form–function relations characteristic of later-language development (Berman, 2007).

The results for narratives point to a slowly progressing increase in the prevalence of noun-modifiers, with the only significant difference emerging between grade 4 and adults. In expositories, noun-modifiers already became significantly more prevalent in high school, and continued to rise even after that point; adult expositories had a mean of 0.58 noun-modifiers per clause – more than one in every two clauses – compared with just 0.28 in high-school expositories. In fact, noun-modifiers were the only type of PP that significantly increased after high school – thus emerging as the major contributors to the growth in overall PP prevalence during that period.

It was hypothesized that noun-modifiers would be more prevalent in expositories – which rely heavily on reference to complex concepts – compared with narratives. This hypothesis was partially corroborated. The proportion of noun-modifiers out of all PPs was indeed greater in expositories. In terms of tokens per clause, however, the only significant genre effect was obtained in the adult group. That is, only in adulthood did expository discourse become a preferred context for employing PPs for the purpose of noun modification.

Recall that participants were instructed to write about ‘problems between people’ – which contains a noun-modifying PP. It is possible that this led to a greater prevalence of noun-modifiers than would have been found otherwise, since this wording was echoed in some of the texts. However, the use of PPs as noun-modifiers was a wide phenomenon, far from being restricted to this specific expression.

A qualitative examination of noun-modifying PPs reveals some further interesting trends. These PPs initially had very restricted uses: in 4th-grade narratives they were almost exclusively recruited to provide simple qualifications of people, e.g.

(12) ye’led  me-ha-kita sheli

boy from-the-class of.1sg

‘a boy from my class’

while in expositories, there were mostly juvenile anaphoric expressions including a sequence of the preposition ke ‘as’ + demonstrative, e.g.

(13) lo lignov ve-dvarim  ka-éle

not steal.INF and-things as-these

‘not to steal and things like that’

Starting from grade 7, noun-modifiers began to be used with much more sophistication, for qualifying people as well as inanimate objects and abstract concepts from various perspectives, e.g.

(14) nórmot  le-hitnahagut xevratit

norms to-behavior social.SG.FEM

‘norms for social behavior’

(15) be’aya  im axat me-ha-xaverot sheli

problem with one.FEM from-the-frienrds.FEM of.1SG

‘a problem with one of my friends’

An especially important use of noun-modifying PPs is the modification of nominalized forms. Nominalization – the construal of situations as static entities – is the bread and butter of expository discourse, which deals with general phenomena rather than the unfolding of specific events, but is also a hallmark of mature narration, which typically adopts a detached, abstract stance (Berman, 2005; Ravid & Cahana-Amitay, 2005). From grade 7 on, but most markedly in adults, PPs supported the use of nominalizations by designating related participants and circumstances, thus maintaining a full event structure around them – in expositories, e.g.

(16) muda’ut  le-be’ayot

awareness to-problems

‘awareness of problems’

(17) vikuxim ba-kvish

disputes on-road.DEF

‘disputes on the road’

as well as in narratives, e.g.

(18) ha-sin’a  beyn shtey ha-kvutsot  hayta

the-hatred between two.FEM.GEN the-groups was.SG.FEM

hadadit

mutual.SG.FEM

‘The hatred between the two groups was mutual’

Other roles

Discourse markers are PPs serving a procedural function, referring to the discourse itself. In expositories, discourse markers were mostly of the form le-da’ati ‘to-opinion.1SG = in my opinion’, or else – mostly in adults – served to provide ‘structure-building instructions’ (Britton, 1994), guiding the reader through the process of constructing the intended network of ideas, e.g.

(19) mi-tsad sheni

from-side second.SG.MASC

‘on the other hand’

In narratives, discourse markers first appeared in grade 7, mostly providing interpersonal evaluatory cues, e.g.

(20) hi lo hayta muxana lehakshiv  bi-xlal

she not was.3SG.FEM prepared.SG.FEM listen.INF in-all

‘She wasn’t willing to listen at all’

Adjective-modifiers and PP-modifiers are the two types of PPs which first emerged in grade 7. Common to both these roles is a tertiary function, as modifiers of modifiers – either adjectives or (mostly verb-adjunct) PPs – allowing the construction of elaborated modifying expressions. PP-modifiers are the least common type in the sample, remaining rare in all age groups. Adjective-modifiers grew in prevalence throughout the age-span investigated, and were generally more prevalent in expositories, where there are simply more adjectives (Ravid & Levie, 2010). Adjective-modifiers mostly conveyed an intensifying meaning, e.g.

(21) xashuv  be-yoter

important in-more

‘very important’

or else served to delimit the attributive scope of an adjective, e.g.

(22) kashe  le-fitron

hard to-solution

‘hard to solve’