Projects 2022

Corpus Evidence for Delineating Constructions

(Third Party Funds Group – Sub project)

Abstract:

CxG and many other usage-based approaches agree that language consists of pre-fabricated form-meaning pairings of varying sizes (e.g. Goldberg 1995, Hunston & Francis 2000, Sinclair & Mauranen 2006, Wray 2008), which are called constructions in CxG. In contrast to approaches that understand language as a probabilistic system, such as lexical priming theory (Hoey 2005) or the EC-Model (Schmid 2020), constructions are usually conceptualised as discrete symbolic units or the “nodes of a symbolic network” (Diessel 2019: 249), possibly emerging from the generalisation of associational patterns or clusters of memory traces (e.g. Goldberg 2019). Prior research is typically focused on extensive linguistic analysis and discussion of a relatively small set of specific constructions (such as the English ditransitive or the let alone construction). Such studies have not been able to establish clear-cut criteria and diagnostics for determining at scale, i.e. with broad coverage, which form-meaning pairings should be considered as constructions and which elements (lexical items, restricted or open slots, and grammatical features) should be included in a given construction. While it is evident in a usage-based approach that there can be no dichotomic distinction of constructions vs. non-constructions4 and that “constructionhood” is a matter of degree, binary decisions on an inventory of constructions still have to be made for the purposes of linguistic analysis and the systematic compilation of a broadcoverage reference constructicon.

First efforts to build such a reference constructicon have been started for different languages, including English (Perek & Patten 2019) and German (Ziem et al. 2019). They build on existing lexical resources such as FrameNet (Perek & Patten 2019) and/or manual in-depth analysis of selected constructions5 (Ziem et al. 2019). Automatic identification of constructions has only been attempted by a small number of exploratory studies, based on word n-grams (Shibuya & Jensen 2015), hybrid n-grams of words and POS tags (Forsberg et al. 2014), or a combination of dependency-based co-occurrence with distributional clustering (Martí et al. 2019). All three studies focus on extracting and ranking construction candidates for manual inspection, but do not discuss identifying criteria or generate additional quantitative evidence for human annotators. Gries (2003) carries out a small feasibility study on finding prototypical instances of a given construction, but does not address the issue of construction identification.

This project explores how and to what extent quantitative data from large corpora can contribute to the task of delineating constructions, i.e. help researchers to assess the degree of “constructionhood” of a candidate construction (CxCand), develop systematic defining criteria for this assessment, and lay the groundwork for (semi-)automatic identification of constructions at scale. The project combines computational big data analysis of English and German corpora with constructicographic work (Lyngfelt et al. 2018), extending the collo-profile approach proposed by Herbst & Uhrig (2019: 177ff) for argument structure constructions. It addresses three central research questions: Q1: Does quantitative evidence from large corpora improve the manual identification of constructions and the development of defining criteria? Q2: What statistical measures are suitable as an operationalisation of such quantitative data, providing a basis for computing an index of “constructionhood” and for the automatic identification of constructions?

Q3: Can context-sensitive neural word and phrase embeddings be used as a corpus-based approximation of construction meaning?
The project starts by extracting large databases of CxCand from English and German Web corpora of more than 10 billion words, based on pre-defined syntactic patterns such as verb argument structure. The extraction relies on an existing HPC infrastructure for parsing large corpora at FAU. Widely-used criteria for determining “constructionhood” such as productivity, compositionality / idiomaticity and schematicity / lexical specificity (Ziem et al. 2019: 69f) are operationalised in terms of corpus frequency, productivity of slots, statistical association between lexical elements, morpho-syntactic preferences, context entropy, etc. They are computed from the CxCand database using state-of-the-art measures from methodological research carried out at FAU, which provide the basis for answering Q2. Following Herbst & Uhrig (2019), the meaning aspect of a CxCand is initially approximated by the collo-profiles of its open slots. A thorough constructicographic analysis of different sets of CxCand sheds light on Q1 (whether constructions can clearly be identified) and Q2 (which quantitative measures are most useful for this purpose). These sets include well-studied examples of constructions from the literature (used for validation of the approach), sets based on a syntactic pattern (such as mono-transitive verb argument structure), and sets based on a lexical item (in particular various prepositions, in collaboration with project #9). The most challenging and open-ended aspect of the project explores the use of context-sensitive word and phrase embeddings (e.g. Devlin et al. 2019) to operationalise the semantics of a CxCand, following the distributional hypothesis (Harris 1954) and recent proposals for a distributional CxG (DisCxG: Rambelli et al. 2019). If successful, i.e., if there is a positive answer to Q3, not only the form of a construction but also its meaning can be studied based on corpus evidence.

Research questions Q1 and Q2 directly address GRQ CON1 (How do we identify constructions? Can they be seen as discrete units?) and GRQ CON2 (To what extent is constructional knowledge determined by collo-profiles? How can we measure the lexical specificity vs. productivity of constructions slots?). An important part of the constructicographic analysis is to delineate between a CxCand and related constructions, such as a generalisation of the CxCand or an overlapping combination of two constructions. In this way, the project also addresses GRQ NET1 (How can computational methods help reveal the network character of constructional space?).

The project will contribute a substantial number of entries to the RCnn, combining constructicographic descriptions with rich quantitative evidence. A suitable representation format for these entries will be developed in close collaboration with the PDR. The CxCand database constitutes a valuable resource for other projects working on English or German constructions; an extension to other languages is envisaged for the second phase of the RTG.

Multimodal Constructional Space

(Third Party Funds Group – Sub project)

Abstract:

When it comes to human face-to-face communication, speakers make use of various modalities to deliver and interpret messages. This complex operation involves not only the verbal exchange of linguistic forms, but also the use of facial expressions, gestures, and prosody. In fact, a number of studies have shown that many gestures and linguistic forms systematically co-occur with one another (Cienki, 2015; Ningelgen & Auer, 2017; Ziem, 2017; Zima, 2017b). Following a cognitive/usage-based model, we know that language learners and users keep track of usage events, and that knowledge of language is constantly shaped and re-shaped with each instance of use (Bybee, 2010). One of the challenges of linguistic theory is therefore to account for these multimodal phenomena.

The main focus of this project is on modeling multimodality in a Construction Grammar (CxG) framework (Goldberg, 1995, 2006, 2019). Over the past few years, there have been several proposals that address these types of phenomena (Cienki, 2017; Herbst, 2020; Hoffmann, 2017; Mittelberg, 2017; Schoonjans, 2017; Turner, 2018; 2020a; 2020b; Uhrig, 2021; Ziem, 2017; Zima, 2017; Zima and Bergs 2017). Still, there are various theoretical and practical aspects yet to be addressed. Some of the discussion points that were brought up concern the theoretical status of multimodal constructions, whether the constructicon is multimodal, and whether a Multimodal Construction Grammar is needed.

In detail, this project is set to investigate three kinds of constructions where multimodal phenomena are observed, and use them as case studies to understand and suggest ways of modeling multimodality, following a CxG framework. These are cases whereby a linguistic form is observed to systematically and frequently co-occur with a gesture as in “I came this close to 🤏 winning the lottery”; a gesture that seemingly has a syntactic role in the utterance; and a gesture where no association is found with particular linguistic forms, but rather is a case of free combination such as air quotes (Uhrig, 2020). All of this will be done in the form of corpus-based studies. The data will be extracted from the large repository of audio-visual data NewsScape English Corpus, and analyzed using a variety of tools such as CQPweb, Red Hen Rapid Annotator, Elan, and Praat (Uhrig, 2021).

By the end of this project, we would like to suggest ways of delineating and modeling multimodal constructions in a CxG framework, account for the type of information that should be considered when describing constructions, which is not a trivial matter, and apply data science methods to multimodal communication research to identify and extract gestures from multimodal corpora, and to use statistical methods to analyze them.

Form and meaning as factors in the identification and learning of constructional slots – English phrasal verbs and verb-preposition combinations

(Third Party Funds Group – Sub project)

Abstract:

While there is considerable evidence to support the claim that linguistic knowledge ‘exists’ in the mind in the form of constructions (as defined in CxG- and usage-based approaches) (Casenhiser & Goldberg 2005; Goldberg 2019; Dąbrowska & Lieven 2005), there is comparatively little research on how particular constructions are stored and processed in the brain. Cappelle, Shtyrov and Pulvermüller (2010) have argued on the basis of mismatch negativity that phrasal verbs (look. up) are processed in the brain like single lexemes and expressed a warning “against a total abolition of a lexicon-syntax distinction” (Pulvermüller, Cappelle & Shtyrow 2013: 415). This project aims to follow up this issue, which is central to the design of a model of constructional space, by obtaining fMRI, MEG and EEG data to investigate whether phrasal verbs (as idiomatic combinations) are exceptional in this respect or whether similar effects can be found in other constructions containing lexical elements – e.g. argument structure constructions with prepositions (decide on) or clauses (say that ..., aim to do) and how these compare with constructions without such lexical elements such as the ditransitive construction (X VERB Y Z).
The results of this research will make a valuable contribution to GRQ1 in that they will provide evidence to resolve the issue of how to delimit particular constructions, slot-fillers of constructions and the nature of chunks with respect to purely lexical associations and chunks with more schematic slots. The analysis will comprise a number of behavioural experiments such as eyetracking and neurolinguistic measurements. The first task of the project will consist in designing (a) a text that embeds a sufficiently large number of tokens of the types of constructions to be explored in a kind of natural narrative as well as (b) a set of isolated test sentences highlighting the same phenomena. Candidates for test items will be identified by means of cluster and frequency analyses of corpora of present-day English. In a second phase, eye-tracking and brain activity measurements will be carried out with ca. 50 native speakers of English in order to explore (i) whether the reactions to phrasal verbs in the test battery confirm the results of P/C/S 2013, and (ii) to what extent they differ from measurements for the other types of constructions identified.
By including running text into the research design, we deliberately go beyond previous research to ensure that subjects’ performance will be studied in a more natural setting than is possible by isolated sentences. This part of the research will be based on methodology developed at FAU enabling multi-modal neuroimaging measurements during continuous speech perception (Schilling et al 2021) and multivariate cluster analysis of the resulting spatio-temporal neural activation patterns (Kriegeskorte, Mur, & Bandettini 2008; Krauss et al 2018; Schilling et al 2021).

Representation and processing of constructions in the brain

(Third Party Funds Group – Sub project)

Abstract:

The main objective of this project is to study the representational and computational principles of construction storage and processing, and their structural and functional implementation in the brain. Recent experimental findings in neuroscience provide converging evidence for the neural and psychological plausibility of network-like representations of hierarchical linguistic structures such as constructions (Herbst 2018a, 2020) on abstract cognitive maps. Cognitive maps are mental representations that serve an organism to acquire, code, store, recall, and decode information about the relative locations and features of objects (Tolman 1948, O’Keefe & Nadel 1978).
Electrophysiological research in rodents suggests that the hippocampus (O’Keefe & Nadel 1978) and the entorhinal cortex (Moser, Moser & McNaughton 2017) are the neurological basis of cognitive maps. There, highly specialised neurons including place (O’Keefe & Dostrovsky 1971) and grid cells (Hafting et al. 2005) support map-like spatial codes, and thus enable spatial navigation (Moser, Kropff & Moser 2008). Furthermore, human fMRI studies during virtual navigation tasks have shown that the hippocampal and entorhinal spatial codes, together with areas in the frontal lobe, enable route planning during navigation (Spiers & Maguire 2006, Spiers & Gilbert 2015, Hartley et al. 2003, Balaguer et al. 2016) based on distance preserving representations (Morgan et al. 2011). Recent human fMRI studies even suggest that these maplike representations extend beyond physical space to more abstract social and conceptual spaces (Epstein et al. 2017), thereby contributing broadly to other cognitive domains (Schiller et al. 2015), and thus enabling navigation and route planning in arbitrary abstract cognitive spaces (Bellmund et al. 2018). Besides contributing to spatial navigation, the hippocampus also plays a crucial role in episodic and declarative memory (Tulving & Markowitsch 1998) by receiving highly processed information via direct and indirect pathways from a large number of multi-modal areas of the cerebral cortex (Battaglia et al. 2011) including language related areas (Hickok & Poeppel 2004).
Finally, some findings indicate that the hippocampus even contributes to the coding of narrative context (Milivojevic et al. 2016). These findings provide a novel theoretical framework of language representation and processing. There, hippocampal coding would enable flexible representational mapping of linguistic structures across a wide range of scales and hierarchical levels, from phonemes, single words and collocations (Evert 2008), through valency patterns (Herbst & Uhrig 2019, Herbst et al. 2004), to idioms and abstract argument structure constructions (Herbst 2018, Herbst & Uhrig 2019).
In particular, the project is to explore the hypothesis that linguistic constructions are represented as multi-scale network-like maps, and that constructions are combined in the process of formulating an utterance by navigating on these maps, i.e. a certain utterance would correspond to a certain route. Furthermore, overlap and blending (Herbst 2018b) of constructions are realized and guided as and correspond to switching between different levels of the multi-scale maps. In contrast to the brain, computational models have the decisive advantage that they are fully accessible, i.e. the temporal evolution of all model variables can be read out for further analysis at any time. In addition, we can perform arbitrary experimental manipulations on these models. Therefore, starting from contemporary machine learning approaches for the representation and processing of natural language (e.g. Devlin et al. 2019), computational models will be constructed that can build and navigate on hierarchical map-like multi-scale representations of language, thereby being capable of representing natural language input, transforming these representations according to pre-defined tasks (e.g. re-phrasing), and producing modified natural language output. Using evolutionary optimization, the biological fidelity of the initially constructed models will be increased iteratively by applying the concepts of variation and selection. In order to select those candidate models that best fit to measured brain activity, internal model activity and brain activity will be compared using an advanced methodology comprising multivariate statistics (Kriegeskorte, Mur & Bandettini 2008, Krauss et al. 2018, Schilling et al. 2021) and Bayesian model selection (Mark et al. 2018). After each selection step, new “child models” will be created by adjusting their architectures and parameters using machine learning approaches such as neural architecture search (Elsken, Metzen & Hutter 2018, Wistuba, Rawat & Pedapati 2019, Liu et al. 2018, Zoph & Le 2016, Chen et al. 2019) and genetic algorithms (Gerum et al. 2020). So far, neuroscientific studies of language have mostly employed over-simplified experimental paradigms, e.g. by focussing on single word processing or sentences in isolation. Very recently, the advantages of using natural, connected language such as narratives for neuroimaging studies have been discussed (Willems, Nastase & Milivojevic 2020, Jääskeläinen et al. 2020, Hamilton & Huth 2020, Hauk & Weiss 2020, Schilling et al. 2021). Therefore, multi-modal (fMRI, MEG, EEG, iEEG and ECoG) measurements during continuous speech perception (listening to audiobooks) will be performed, as described in detail in our preliminary work (Schilling et al. 2021).

Representation and acquisition of agreement relations in a usage-based framework

(Third Party Funds Group – Sub project)

Abstract:

In most grammatical frameworks, agreement relations such as subject-verb and adjective-noun agreement are handled by using abstract features and a formal operation which copies or unifies them. Such an approach, however, is incompatible with Langacker’s (1987) content requirement, (which prohibits meaningless grammatical features), raises numerous learnability problems, as well as being problematic for empirical reasons (Kibrik 2019). Building on prior work in construction grammar and cognitive linguistics (Acuña-Fariña 2018, Kibrik 2019, Miorelli & Dąbrowska under review), this project will provide a comprehensive inventory of agreement constructions in a particular language (to be agreed with the supervisors) and an explicit account of how these constructions can be learned from the input available to children which makes testable predictions about acquisition or processing, and test these predictions experimentally.

The first phase of the project will involve analysing spontaneous speech data from the CHILDES corpus in order to examine the suggestion made by Miorelli & Dąbrowska (under review) that children could, in principle, attain a high degree of accuracy on agreement in production without any knowledge of agreement features, by simply superimposing lexically specific chunks with partially overlapping semantic and phonological specifications. Consider, for example, the utterance dove è andato il pomodoro? 'where did the tomato go?' produced by a two-year-old Italian child. The utterance contains three agreement relations: between the subject noun and its determiner (il~POMODORO), between the subject and the auxiliary (POMODORO~è), and between the subject and the past participle (POMODORO~andato). In line with earlier studies of production using the so-called traceback method (Lieven et al. 2003, Dąbrowska and Lieven 2005, Dąbrowska 2014), such an utterance can be derived by superimposing the lexically specific unit il pomodoro 'the tomato' and the schema dove è andato il NOUN-o? 'where did the NOUN go?' (which could be derived by generalizing over similar utterances in the input, e.g. dove è andato il ragazzo? 'where did the boy go?', dove è andato il libro? 'where did the book go?', etc.).

If this account is correct, knowledge about agreement is (at least initially) 'hidden' inside other constructions, which raises interesting theoretical questions about the nature of speakers' knowledge of agreement. This part of the project will address the GRQs CON1 (How do we identify constructions – in this case, how do we know when/if a speaker has mastered a particular agreement construction, and how general is it?) and USE3 (How are constructions combined in the process of formulating an utterance and what role do co-occurrence, overlap and blending play in this process? – cf. Dąbrowska and Lieven 2005, Herbst and Hoffman 2018).

In the second phase, specific hypotheses about the acquisition and processing of agreement relations will be tested using elicited imitation and grammaticality judgment tasks with sentences containing agreement violations. In particular, we will examine the suggestion made above that young children's knowledge of agreement is largely 'buried' inside chunks (i.e., it is not an not independent construction), and a hypothesis derived from Acuña-Fariña's work, namely that young children's knowledge of agreement is to a large extent based on purely phonological patterns (il …o …o, as in il ragazzo alto 'the tall boy' and la …a …a, as in la camicia bianca 'the white shirt'). It is anticipated that, across development, speakers will be better at detecting or correcting errors when these occur inside fixed chunks (i.e., we anticipate better performance on la camicia *bianco 'the white shirt' than on la camicia *grigio 'the grey shirt') and when they involve phonologically regular patterns like those mentioned above (better performance on la barca *bianco 'the white boat' than on la nave *bianco 'the white ship'). This part of the project will address GRQ ENT1 (How do frequency and salience influence entrenchment?). If time allows, we will also conduct ERP studies to examine the brain response to these types of agreement violations and thus address ENT3 (To what extent do measures of neural activity during language processing coincide with the results of behavioural and corpus data and how does this expand our understanding of how constructions are stored and processed in speakers' brains?).

Representation and acquisition of idiomatic constructions in L1 and L2 learners

(Third Party Funds Group – Sub project)

Abstract:

Although the study of idioms has played a crucial role in the development of Construction Grammar (e.g. Croft and Cruse 2004, Fillmore et al. 1988), there are still only a few studies comparing constructions with a relatively high degree of idiomaticity in different languages (e.g. Abel 2003, Apresjan 2014, see also Gries and Wulff (2005) for a more general study of constructions in foreign language learners). Idioms that are motivated semantically are common across different languages (e.g. Dobrovol’skij and Piirainen 2010). These idioms can share the same underlying functional (semantic/pragmatic) properties and form such as A wolf in a sheep’s clothing in English, Ein Wolf im Schafspelz in German and gorgī dar lebāse mish in Persian. Or they can be motivated by similar symbolic or cultural concepts or coercion but have different forms as in to take the bread out of someone’s mouth in English, das ist ein hartes Brot in German, and nān-e kasi rā ājor kardan ‘make someone’s bread a brick’ in Persian. he major goals of this project are (i) to examine how and to what extent the processing of L2 idiomatic constructions is influenced by the existence of similar idiomatic constructions in the L1 and (ii) to determine which factors, e.g., age of learning, typological similarity or frequency and context of use, contribute to the entrenchment of L2 and L1 idiomatic constructions.
In the first phase of the project, a contrastive corpus analysis (e.g. Granger 2002) will be used to identify semantically similar idiomatic constructions and their frequencies in the three languages: English, German and Persian. By using phrase classification tasks (e.g. Swinney and Cutler 1979) and priming experiments (including sentence generation and completion tasks), similar to the ones, for example, described by Sprenger et al. (2006) and Yeganehjoo and Thai (2012), the receptive and productive knowledge of comparable idiomatic expressions in L1 and L2 learners of English, German and Persian will be tested in order to determine to what extent the processing of L2 idioms is facilitated by the existence of similar constructions entrenched in the L1 of an L2 learner. This part of the project will address GRQ CON4 (To what extent can constructions (and their constituents) identified in one language be equated with superficially similar constructions in another language?) as well as CON1 (How do we identify constructions?).
The results of the experiments obtained in the first phase will form the basis of the analyses carried out in the second phase, in which specific hypotheses about factors possibly influencing the entrenchment of L1 and L2 idiomatic constructions will be tested. In particular, by comparing the results obtained for different age groups of L1 and L2 learners, different L1 and L2 pairings of typologically more similar and more distinct languages, different types of idiomatic expressions (i.e., those that share the same functional properties and form across different languages vs. those that are motivated by similar symbolic or cultural concepts but have different forms) and idiomatic expressions differing in terms of their frequency and contexts of use, the possible effects of different factors that have generally been claimed to affect the processing and entrenchment of constructions will be examined with regard to the specific role they play in the processing of L1 and L2 idiomatic expressions (for the possible contribution of different factors to the processing and entrenchment of constructions and/or idiomatic expressions, see, e.g., Abel 2003, Apresjan 2014, Divjak and Cardwell-Harris 2015, Steinkrauss and Schmid 2016, Wasserscheidt 2014). This part of the project will address GRQ ENT1 (How do factors such as frequency, salience, dispersion and age of acquisition influence entrenchment?). If time permits, we will also conduct ERP studies to examine the neural correlates of L1 and L2 idiomatic constructions in L1 and L2 speakers, which would allow us to address ENT3 (To what extent do measures of neural activity during language processing coincide with the results of behavioural and corpus data and how does this expand our understanding of how constructions are stored and processed in speakers’
brains?). Naturally, insights gained in this project will also make a contribution towards applying Construction Grammar to foreign language teaching (De Knop and Gilquin 2016, Herbst 2017, Erfurt & De Knop 2019, see also special issue of ZAA edited by Piske, Herbst & Uhrig 2014).

Artificial language learning as a window to the early entrenchment of constructions

(Third Party Funds Group – Sub project)

Abstract:

This project examines the roles of frequency and semantic transparency on the learning and entrenchment of constructions in first and second language acquisition. First, in a series of studies, we focus on the comprehension by child and adult native speakers of Spanish of different variant forms of object relatives in Spanish (e.g., el niño (al que) el abuelo abraza 'the boy that the grandpa is hugging') that differ in terms of absolute frequency in oral discourse (Butler, 1992) and transparency of grammatical role assignment (Slobin, 1973). Crucially, this focus on highly similar syntactic structures not only allows for putting the two factors of interest in opposition in a controlled manner, but also provides experimental evidence informing the debate about how to determine what constitutes one (and the same) construction within a CxG framework. A follow-up of this experimental paradigm including word order variations in Spanish object relatives allows us to further explore the roles of additional morphological markers and primary cue timing in syntactic ambiguity resolution.

Secondly, we resort to an artificial language learning paradigm (e.g., Pili-Moss, Brill-Schuetz, Faretta-Stutenberg & Morgan-Short 2020; Rebuschat, Monahan & Schoetensack, 2021) in order to be able to systematically manipulate linguistic factors during the earliest stages of L2 acquisition. To this end, we expose native English speakers to an artificial language based on Portuguese phonotactics, manipulating case marker salience in terms of syllabicity, stress and sonority (high: /-'ka, -'ʃi/; low: /-l, -ɾ/). Participants are then grouped into two conditions where the marker has either high or low phonetic salience. This will provide further insights on the role of phonetic salience in the emergence of constructional knowledge and in the acquisition of non-native morphology, particularly,agent and patient markers. In sum, this project assesses the relative influence of factors that are known to separately facilitate the acquisition of constructions, contributing thus to our understanding of the factors that shape constructional entrenchment and to the discussion around the concept of construction itself.

German verbs with particles or prefixes in language change: Form, meaning, and syntax

(Third Party Funds Group – Sub project)

Abstract:

Word formation processes occupy a central position in constructional space in that they involve units which are small in size but nevertheless complex and because word formation processes are located between the lexical and syntactic poles of this space (cf. Felfe 2012, Michel 2014). However, studies on word formation in language change are rare. For this reason, we will address one key aspect of the diachronic emergence of verbs with particles and prefixes in this project. In the first phase, German verbs with durch-, hinter-, über-, um-, unter-, and wider- will be analysed with respect to their formal and semantic change from initially loose syntagms to stable word units. The change in the form-function pairs will be described alongside the variation and change of the syntactic constructions in which they are embedded. In accordance with Booij’s view (2010: 3) that word formation patterns are “abstractions over sets of related words”, and that complex words are based on constructional schemata, we are aiming to identify different levels of abstraction for the combinations analysed. In the second phase, we want to uncover the multifacetted network of the continuum of schematicity in diachrony by tracing the processes of idiomatization resulting in non-transparent expressions. In the third phase, the conditions of the spread of word formation change phenomena in the language will be analysed. Methodologically, this is an empirical research project based on the analysis of historical language corpora of
German. In addition to historical word formation research, approaches of Construction Grammar and Relational Morphology (Jackendoff & Audring 2020), grammaticalization research and valency grammar are drawn upon.
The first phase of the project will involve analysing data from historical language corpora of German such as the following: Deutsch Diachron Digital (DDD) Altdeutsch, Mittelhochdeutsch, Frühneuhochdeutsch, Mittelhochdeutsche Begriffsdatenbank, GerManC, Deutsches Textarchiv (DTA) and DWDS. The analysis will include the description of all verbs with durch-, hinter-, über-, um-, unter-, and wider- to be found in these corpora. The development of the selected particles or prefixes will be studied phonologically and graphemically to determine grammaticalization processes (erosion, univerbization) diachronically. All of these verb combinations will be described (and annotated, wherever possible) morphologically (word formation base, separable particle or non-separable prefix, etc.) and syntactically (valency, semantic role, lexical filler of slots, syntactic collocations, topology). Subsequently, the individual verbs will be assigned to semantically defined schemas (e.g. ‘local’, ‘aspectual’) and subschemata defined by their specific word formation components and their syntactic environment. This part of the project will focus on the GRQ CON1 (How do we identify constructions – in this case, to what extent do word formation schemata differ from lexical items, and what is construction change in word formation?; cf. Hilpert 2013, Jackendoff & Audring 2020).

The second phase of the project will deal with constraints on open slots in syntactic use and their semantic fixation on certain words: for example, in the construction x geht um, literally ‘x walks around’, the x slot was highly productive in MHG, whereas in NHG it is restricted to ein Gespenst ‘a ghost’, eine Seuche ‘a plague’, eine Liste ‘a list’ and a small number of related NPs. We will model the processes and degrees of idomaticity in meaning change in terms of a "continuum of schematicity" (Croft & Cruse 2004: 255). The data for different periods will be analysed synchonically before a diachronic analysis is attempted – cf. Habermann (in press). This part of the project will address the GRQs CON2 (How can we operationalize the degree of lexical specificity vs. productivity of construction slots?).

A third aim of this research will ask to what extent some factors favour the spread of word formation constructions from individual use in particular situations, text types, or regions to more general use throughout society (as far as this can be seen historically). To find an answer to this question, the verbs of the text corpus will be described (and annotated) for frequency, with special attention paid to hapax legomena, taking into account factors such as individual usage (author), stylistic (text type) and regional context (see also Bybee 2015). (This latter aspect could be extended in a follow-up project in the second phase of the RTG.) This part of the project will address GRQs ENT1 (How do frequency, salience and dispersion influence entrenchment?) and USE4 (How do the factors communicative intentions, socioeconomic status, and dialect result in language change at the community level at different timescales?).

Functions and cognitive semantics of prepositions in complex constructions

(Third Party Funds Group – Sub project)

Abstract:

Fundamental work in cognitive linguistics has highlighted the role of space and spatial expressions to the cognitive organization of language (Talmy 1983, 2008, Lakoff and Johnson 1980), which resulted in a very detailed interest in the semantics of prepositions and their organization in semantic networks already quite early on (see Sandra and Rice 1995 for an overview and a critique). This project will be concerned with an analysis of German prepositions and the central role they play in the expression of space, time, instrumentality, and modality both in concrete and more abstract uses such as the so-called governed prepositions (see e.g. Breindl 1989 for a detailed account of prepositional objects). Recent constructional treatments of German prepositions, such as the ones by Rostila (2014, 2015, 2018) and Zeschel (2019) will provide the theoretical starting point. German constructions with prepositions such as mit ‘with’, unter ‘under’, zwischen ‘between’ and über ‘over’, um ‘round’, zu ‘to’ form semantic nests of similarity, which express roles such as PARTNER or TOPIC in the communication frame (e.g. with diskutieren ‘discuss’, sprechen ‘speak’, Diskussion ‘discussion’, Debatte ‘debate’) (compare the families of overlapping constructions with the preposition into, Herbst & Uhrig 2019). The aim of the project is to provide a corpus-based description of argument structure constructions with these prepositions and an illustration of the way they cluster, i.e. of overlap or links between the various constructions postulated, making use of semantic frames (e.g. German FrameNet) or image schemata. These descriptions, which will also consider aspects such as text types, cultural background etc., can then become entries of the general research constructicon, which is one common aim of the RTG. This part of the project addresses CON 1. (CON1: How do we identify constructions (what are their defining criteria; are they better seen as discrete units, prototypes, attractors in constructional space, or nodes in a network of cognitive associations)?) A contrastive analysis between selected German and English prepositions (in the spirit of Uhrig & Zeschel 2016) will be carried out to determine the extent of language-specific encodings (CON4: To what extent can constructions (and their constituents) identified in one language be equated with superficially similar constructions in another language?). This is directly related to the question of the degree of detail and item-specificity with which such prepositional constructions should be distinguished and stored in the mental constructicon and the reference constructicon (CON2: To what extent is constructional knowledge determined by the specific items occurring in them (colloprofiles) and how can we measure and operationalize the degree of lexical specificity vs. productivity of construction slots?), because if it turns out that uses across English and German are not predictable, a stronger role of storage will have to be assumed. More specifically, possible factors determining construction status will be investigated, including variables related to diachronic or regional variation, and weighted against factors such as individual differences (Dąbrowska 2012a, 2012b, 2015b) and socio-cultural conditions of the use of a construction (USE1: What factors influence speakers’ choices from a range of competing constructions?). The project will make use of various methodological approaches, including hermeneutic analysis of meaning and semantic similarity supported by judgments tests. Most importantly, however, the project will build on the corpus-analytic procedures described by Schierholz (2006) and Zeschel (2015) for the monolingual research and Uhrig & Zeschel (2016) for the contrastive aspects.

Argument structure constructions with prepositions and phrasal verbs in Arabic and other Semitic languages

(Third Party Funds Group – Sub project)

Abstract:

One of the key challenges of this project is to explore to what extent and in what form the Construction Grammar approach can be applied to a Semitic language. Indeed, it would seem that the concept of constructional space is apt to capture and formally analyze the form-meaning pairings entailed in the system of diatheses/binyanim expressing voice, intensity, reflexivity, interaction, and other qualities (cf., e.g., Retsö 1989) characteristic of Arabic and other Semitic languages, but, to the best of our knowledge, Construction Grammar has not been used in the analysis of Arabic so far. The main focus of the present project will lie on verbal constructions with prepositions and particles, and in particular on the interaction of prepositional constructions with verbs in the creation of meaningful units (which in turn may be stored holistically by speakers as low-level constructions) and on phrasal verbs in Arabic and Semitic in general.
Even a superficial look at an Arabic-European lexicon reveals various and seemingly contradicting meanings for the verb daʕā ‘to call’, namely ‘to pray’ and ‘to curse’. A closer look shows that these meaning depend on the prepositions governed by the core verb, li ‘for’ and ʕalā ‘upon, against’, respectively. Diachronically, daʕʿā li means ‘to invoke God on someone’s behalf’, and daʕā ʕalā ‘to invoke God against someone’, resulting in the contradictory semantics.
Synchronically, however, speakers (usually) are not aware of this background. Rather, they use the combination of core verb and preposition as an exocentric construction. This type of exocentric constructions permeates the whole verbal system of Semitic and still awaits a systematic investigation. Certain prepositions in this context have undergone grammaticalization (cf., e.g., Rubin 2005). By and large, the same holds true for the meaning of derived diatheses (binyanim) in Semitic and the relationship between intransitive, transitive, and ditransitive uses of one and the same verb in different constructions. While the meaning of derived diatheses is predictable in some cases, this is not the case in other cases, which, again, can be better captured in terms of exocentric constructions used morpho-syntactic building blocks. Experiencer constructions also play an important role in this context (cf., e.g., Retsö 1987 and Edzard 2016). A related type of construction in Semitic is presented by phrasal verbs. There is a host of constructions in Arabic and Semitic, among them cognate object constructions which reflect the concept of “conflated complements” (e.g. Talmy 1985). What is more, there also exist constructions, notably in Ethio-Semitic, that typically consist of an invariable element (e.g. an ideophone, onomatopoetic element or noun, sometimes with an opaque meaning) and a grammaticalized (semantically bleached) verb with the original meaning ‘to say’ or ‘to do’. In Amharic, the relevant verbs are ʔalä ‘to say’ and ʔadärrägä ‘to do’. Examples include k’uʧʧ’ ʔalä ‘to sit’, zɨmm ʔalä ‘to be quiet’, k’uʧʧ’ ʔadärrägä ‘to put down’, and täsfa ʔadärrägä ‘to hope’. The semantics of zɨmm ʔalä ‘to be quiet’ (lit. “to say zɨmm“) perfectly illustrates the relevance of the concept of construction.
The project is to explore the hypothesis that a model that combines the level of rather abstract argument structure constructions (Goldberg 1995, 2006) with the valency properties of particular lexical units by regarding lexical items as an integral part of these constructions (Herbst 2018, 2020; Goldberg & Herbst 2021) would provide an adequate framework for describing valency phenomena in Semitic languages.
The combination of verbs and prepositions on the one hand, and the internal composition of phrasal verbs in Semitic on the other hand, often represent an exocentric scenario, i.e. the meaning of the whole verb phrase cannot be automatically deduced from the meaning of the constituents; in other words, such cases are constructions (in the sense of Langacker 1987, Goldberg 2019, and Hilpert 2020) par excellence. For Arabic and Semitic, this question has never been systematically analyzed. The clear delineation of the relevant exocentric constructions thus addresses GRQs CON1 (How do we identify constructions?) and CON2 (To what extent is constructional knowledge determined by the specific items occurring in them (collo-profiles) and how can we measure and operationalize the degree of lexical specificity vs. productivity of construction slots?). What is more the GRQs USE1 (What factors influence speakers’ choices from a range of competing constructions?) and USE2 (To what extent do the factors determining the choice of construction differ between speakers with respect to their individual backgrounds and personalities?) are of high relevance especially in the realms of dialectal variation, accommodation to other peoples and cultures, and multilingualism. Furthermore, the constructions analyzed can be included in the RTG’s research construction as examples of how the format for constructicon entries originally developed for English can be used for Semitic languages. The subproject can be linked in a meaningful way with those subprojects that methodologically address questions of valency and the internal composition of verbal phrases, notably the subprojects of Stefan Evert and Thomas Herbst (“Delineating Constructions”) and of Mechthild Habermann and Ludwig Fesenmeier (“German verbs with particle or prefix in language change”).

Constructions beyond the sentence: text-structuring in (esp.) sixteenth-century historiographical texts

(Third Party Funds Group – Sub project)

Abstract:

This project aims to apply the constructionist approach to units larger than the sentence, as suggested by Hoffmann & Bergs (2018) (see also Hoffmann 2015 and Hoffmann & Bergs 2015), in particular in the analysis of sixteenth-century Italian historiographical texts such as Machiavelli’s Istorie fiorentine (1525) and F. Guicciardini’s Storia d’Italia (1561). The textstructuring devices employed in these texts have not received much attention in previous research, since analyses tend to rely on modern editions, in which these texts appear typographically subdivided into smaller units, i.e. into chapters and paragraphs. These subdivisions, however, had only been introduced in 19th century editions, while the original versions, i.e. the 16th century prints, were nearly completely devoid of such typographic textstructuring devices. Nevertheless, the analyses proposed so far (cf. e.g. Blumenthal 1980, Nencioni 1984, Dardano 2017: 282–371) give reason to assume very close relationships between linguistic forms and text structuring functions, i.e. originally, text structuring (foreground vs. background; narration vs. comment; discourse-topic shift etc.; cf. Fesenmeier & Kersten 2018) seems to be expressed by certain recurrent lexicogrammatical patterns, which vary however considerably in size and complexity, for example sentence-initial ma ‘but’ without any adversative value, anaphoric coniunctio relativa-constructions, verb subject ordering, complex hypotactic structures with different types and degrees of subordination.
Traditionally, such lexicogrammatical patterns have been described as stylistic devices and often treated independently from one another. However, at least some of them seem to be related (e.g. coniunctio relativa + subordination + passive + present tense: [Le quali cose] [mentre che] … [si trattano]); moreover, the relation between the “grammatical” elements (determiners, subordinating conjunctions) and the “lexical” elements (e.g. encapsulating noun phrases such as cosa/cose ‘thing(s)’ which function “as a resumptive paraphrase for a preceding portion of a text” (Conte 1996: 1)) of such patterns does not always seem to be grasped in a satisfying manner.
Since one of the advantages of Construction Grammar as a theoretical framework is that the mechanisms developed to describe standard syntactic phenomena can be extended to higher levels of linguistic organization (cf. e.g. Östman 2005, Masini 2016: 75–78, Groom 2019, Hoffmann & Bergs 2018) and since previous work has clearly shown that linguistic conventions also exist at higher levels of linguistic organization, e.g. complexes of clauses revolving around the same discourse topic (cf. Nir & Berman 2010), it seems reasonable to assume the existence of constructions that function as schematic frames for the organization of discourse and whose details (grammatical structure, lexical elements etc.) can be described in a systematic way (CON1: How do we identify constructions – in particular: what are their defining criteria?).
Since Machiavelli’s Istorie fiorentine and Guicciardini’s Storia d’Italia present highly complex syntactic “architectures”, it seems promising to analyse both texts in the analytical framework of CxG, in particular with recourse to the concept of “clause packages”, i.e. “text-embedded units of one or more clauses connected by abstract linkage relations” (Nir & Berman 2010: 748, our italics). Following Berman & Nir-Sagiv (2009: 160), parameters of a more fine-grained analysis could be the number of clauses attached to a main clause, the different types of subordinate clauses, their ordering (in particular with respect to the main clause), and the overall structure (parataxis, hypotaxis etc.). As Machiavelli and Guicciardini strongly differ in their views on both history and historiography, it can be expected that such analyses reveal important differences in terms of clause packaging strategies, differences which in turn should reflect certain “epistemological” differences, just as “the epistemologies and phraseologies of academic disciplines” turned out to be “mutually constitutive” (Groom 2019: 315). The project will thus address GRQs USE1 (What factors influence speakers’ choices from a range of competing constructions?) and USE2 (To what extent do the factors determining the choice of construction differ between speakers with respect to their individual backgrounds and personalities?).
In the first stage of the project, the focus will be “synchronic”, i.e. it will involve an in-depth analysis of the two Italian 16th century texts (thereby applying the CxG framework to a “text-language” in the sense of Fleischman 1991: 252 n. 1, i.e. to a “dead language (langue de corpus), one for which all evidence derives from texts”). Nevertheless, it seems appropriate to also include a “diachronic” perspective by taking into account earlier/later historiographical texts in order to shed light on changes in clause packaging and organization of discourse (cf. the evidence given in Colussi 2014); this relates to GRQ USE4 (How do the factors mentioned in USE1 and USE2 result in language change at the community level at different timescales?). Furthermore, since both the Istorie fiorentine and the Storia d’Italia were translated into French in the 16th century, a contrastive analysis could equally allow for relevant insights in the (construction?) status of previously identified lexicogrammatical patterns in the original texts, since 16th century French does not display the same syntactic devices that can be found in 16th century Italian; therefore one might expect different recurring lexicogrammatical patterns in the two languages.

Argument structure constructions in language contact: intransitive motion in Anglo-Norman

(Third Party Funds Group – Sub project)

Abstract:

Constructionist perspectives have been applied to phenomena of language contact and multilingualism only recently (e.g. Hilpert & Östman 2016, Boas & Höder 2018). A central idea here is the multilingual constructicon (instead of assuming separate systems for each language) which comprises both language-specific and language-unspecific constructions, as put forward, for instance, in Höder’s Diasystematic Construction Grammar (2018) in which contact-induced change is modelled as constructionalization, constructional change, and reorganization in the multilingual constructicon. 

The contact language to be investigated here is Anglo-Norman, i.e., the variety of French used in Britain from after the Norman Conquest to the early 15th century. Anglo-Norman developed several characteristics in which it differed from continental varieties of medieval French. Many of these features can be seen as effects of contact with Middle English, as is clearly visible at the phonological level. Anglo-Norman syntax, by contrast, has been argued to remain largely unaffected by language contact until the mid 14th century (Ingham 2012). This has been shown to be the case mostly for abstract syntactic phenomena such as V2, null subjects, or the clitic vs strong form distinction in object pronouns (Ingham 2012, 2010). Argument structure constructions, however, being more ‘meaningful’, might be expected to be more likely to change towards being less language-specific, particularly in a setting where many verbs are used in both languages anyway, and hence language-unspecific (cf. e.g. Durkin 2014 for the large scale borrowing of French lexis into Middle English). According to Schauwecker (2017), for instance, Anglo-Norman develops a resultative construction with legal speech act verbs, copied from Middle English (à la sentence someone to prison).

One of the aspects in which medieval English and medieval French differ with regard to argument structure is the expression of intransitive motion (Huber 2017, Schauwecker & Trips 2018): Middle English often combines directional prepositional phrases and adverbs (“satellite-framing”) with manner of motion verbs (e.g. ride into the forest) or even non-motion verbs (e.g. toil into the forest), whereas medieval French avoids such combinations. Initial research points towards an Anglicization of Anglo-Norman motion expressions: Huber (in press) shows that the non-motion verbs travailler ‘toil’ and labourer ‘toil’ are attested in motion uses in Anglo-Norman since the late 13th century and Schauwecker’s analysis of four selected manner verbs finds them combined with directional complements (PPs with à and sur) more frequently in 12th to 14th century Anglo-Norman material than in continental French (Schauwecker 2019: 60–61) (cf. also Schøsler (2008: 207) who suspects directional adverbs to be more frequent in Anglo-Norman than in continental medieval French, and cf. more generally the work of the DFG-project BASICS (Stein & Trips, 2015–2021)).

The aim of the project proposed here is to investigate motion constructions in Anglo-Norman in more detail, to find out to which degree these are influenced by contact with Middle English, and whether contact influence is felt earlier on the level of argument-structure-constructions than in more abstract syntactic characteristics of Anglo-Norman. This will be done by analyzing motion expressions in the Anglo-Norman textbase (c. 3 million words, various genres), the Anglo-Norman Yearbooks Corpus (c. 1.5 million words, narrative and dialogical sequences from court hearings) and perhaps the PROME database (c. 8 million words, trilingual parliament rolls) and other, not yet digitized Anglo-Norman texts (editions by the Anglo-Norman Text Society). The project addresses the following GRQs:

- NET3 In multilingual speakers, are the different languages represented in different networks or one multilingual network? In particular, are constructional changes towards the Middle English model predominantly found with verbs used in both languages (e.g. gallop/galoper, hasten/haster), and hence happening on the level of the verb, or are we dealing with changes on the more schematic level of the argument-structure-construction?

- USE4: How do the factors mentioned in USE1 and USE2 [here: multilingualism] result in language change at the community level at different timescales? Particularly [also related to USE3]: if the intransitive motion construction in Anglo-Norman is undergoing change to become more like the Middle English one, does this happen “sneakily” (De Smet 2012), i.e. first in more inconspicuous contexts (coordination with other motion verbs, perfect construction (resultative), reflexive pronoun)?

Multifunctionality in Haitian Creole: New insights from a Construction Grammar perspective

(Third Party Funds Group – Sub project)

Abstract:

Basic typological features of Creole languages include the so-called “multifunctionality” of linguistic items (Lefebvre/Lorange 2015:359). Multifunctionality is generally understood as "the capacity of a linugistic unit to fall within more than one linguistioc class or category" (Véronique 2020:197–198, our translation). For example, Haitian Creole words such as manje or malad are invariable and can be used as predicates in li manje ‘(s)he has eaten’ or mwen malad ‘I’m sore’, but also as subjects or objects in li achete manje li ‘(s)he bought his/her food’, or malad mwen an geri ‘my sore has healed’. Likewise, personal deictic expressions such as mwen ‘1st sg’ or li ‘3rd sg’ can function as a subject, an object, or a possessive marker, and in the case of yo ‘3rd pl’ as a plural marker. These different uses can clearly be distinguished by distributional criteria, but are in each case related to the same semantic core, i.e. eating and soreness (manje, malad), grammatical person and number (mwen, li, yo etc.). In previous research, Haitian Creole word forms that can occur in different syntactic slots have generally been analysed in terms of homonymy. For example, Valdman’s (2007) reference dictionary lists different numbered senses for each of the functions mentioned above, which are assigned to parts of speech such as verbs, nouns, adjectives, and prepositions. However, this treatment of multifunctional items may not only obscure an adequate description of Haitian Creole grammar, but also be an artefact of well-entrenched European grammaticographic traditions, which are not necessarily appropriate for describing languages with other typological features (Broschart 1997:124). Similar observations have been made for English concerning words such as since, before etc. (which are often assigned to the word classes of adverb, conjunction and preposition in dictionaries) or both, this etc. (classified as determiners, pronouns and adverbs) (see e.g. Huddleston & Pullum 2002, Herbst & Schüller 2008). It would thus seem appropriate to develop a model for the word classes in Haitian Creole along the lines of Croft’s (2001) Radical Construction Grammar, in which word class categories are defined on the basis of their occurrence in constructional slots characterized by semantic roles and are seen as construction- and thus language-specific (Croft 2001:106; 2016:383). Word classes can then be regarded as generalizations over usage experiences (Vartiainen 2021:231). Previous studies on English adjectives (Croft 2016, Vartiainen 2021) have shown that the Construction Grammar approach complies only partially with the canonical understanding of parts of speech. What appears to be one word category can often be described as a cluster of constructions that show different frequencies and in which particular items participate to different degrees, so that membership seems to be a matter of degree rather than a categorical property. In addition, the number of word classes to be distinguished varies according to the level of abstraction chosen for the description. Thus, the concept of “word class”, from which “multifunctionality” is implicitly derived, raises a number of questions that are directly relevant to the organization of the constructicon (see also the discussion in Croft 2001:107ff). This makes Haitian Creole a particularly promising candidate to explore the nature of constructions and constructional networks. Against this backdrop, the aim of this project is to identify, systematise and analyse families of constructions in which a selection of Haitian Creole items can occur, and in which they cannot. In order to do this, we rely on a corpus-based approach (cf. also Fitzgerald 2020), using the Corpus of Northern Haitian Creole (Indiana University, Bloomington, ca. 200,000 tokens). A list of “multifunctional” items will be extracted from Valdman’s (2007) dictionary (i.e. items with senses assigned to different word classes). These items will be searched in the corpus based on their surface form (to ensure high recall) and those with at least 50 occurrences will be retained for further analysis. Adressing the GRQs CON1 and CON2, the examples will be analysed and divided into categories, based on common distributional and semantic properties. Items that can be used in similar sets of constructions will be grouped together. On this basis, the following issues relevant to NET2 will be discussed: What are the candidates for constructions that define word classes in Haitian Creole? At which level of abstraction can they be described? How do they relate to each other in terms of polysemy, subpart, metaphorical extension or instance? Can membership be determined in absolute or gradual terms? With how many word-class defining constructions do so-called multifunctional items occur, so that they should be called multifunctional? If this is the case: Can these word class-defining constructions be brought together at a higher level of abstraction? Is the concept of multifunctionality as applied to Haitian and other Creole languages empirically justified under a Construction Grammar framework?