Abstract
The processing of multi-word units and complex words has been one of the main issues of psycholinguistic research in the last decades. However, there is still no mutual consent on how multi-word units, complex words, and their internal constituents are accessed in language processing. Current models of linguistic theory and language processing generally assume that there is no interconnection between the morphosyntactic information of a lexical unit and its phonetic realization. Recent studies challenge this assumption and suggest a relationship between the morphosyntactic, lexical, and pragmatic information of specific lexemes or morphemes and the phonetic signal. The present study adds to these current studies in psycholinguistics and morphophonetics by investigating the French preposition de ‘of’ as a constituent in different construction types. While de occurs regularly as a free lexeme in syntactic structures, it also appears as a bound constituent in lexicalized and grammaticalized constructions. First, this study presents an analysis of French de in eye-tracking data from a reading task with French native speakers. Second, this study presents a statistical analysis of acoustic durations of de from an experimental reading task. The results suggest that the constituent de shows certain peculiarities in its processing and acoustic realization as a constituent in a certain construction type. The results are discussed with regard to current theoretical approaches to the processing of multi-word units, n-grams, and complex words.
1 Introduction
The processing of complex words, such as compounds, has long been the focus of psycholinguistic research (e.g., Rumelhart and McClelland 1986; Taft and Forster 1976). Most of the research on morphological processing of complex words has focused on nominal compounds, such as NN compounds, as in tree house or bus driver. A central focal point is the question of whether complex words are decomposed during processing or whether they are perceived as compositional units. Are complex words stored and accessed as a unit in language processing or are they computed by means of the grammar? What role do the constituents play? Are complex words first perceived as a single unit or are the individual constituents perceived before processing the complex word? The present study attempts to add to the research on the morphological processing of constituents of multi-word units by providing an explorative reading study on the processing and acoustic realization of the French prepositional constituent de ‘of’ in different kinds of multi-word units and as a free preposition in syntactic structures. Up to now, psycholinguistic research has traditionally focused on lexical word classes, such as nouns, verbs and adjectives. Prepositions as a word class are very heterogeneous without a clear definition or delimitation from other word classes. In addition, the status of prepositions is highly controversial in current research, with some defining prepositions as lexical units with a clear semantic meaning and others defining them as grammatical units that merely fulfill linking functions without carrying any semantic meaning. Hence, the present study aims to contribute to this discussion by offering experimental data on the processing of French de in different morphosyntactic environments.
Furthermore, in recent years, new interdisciplinary research areas have emerged, combining phonology and phonetics with morphological research. These research areas, also known as morphophonetics and morphophonology have produced various studies on questions centering on the phonological reality and phonetic realization of morphological units, such as affixes and other types of morphemes coding morphological information. So far, morphophonetic approaches have mostly been neglected in Romance linguistics. Therefore, the aim of the present study is to investigate the interrelation of morphological functioning and phonetic realization in French. One main issue in morphophonetics is to what extent the acoustic signal contains morphosyntactic information and how this information influences the realization of morphemes and lexemes. In current models of language processing, phonetic information has generally been considered to be disconnected from other levels of speech processing (e.g., Levelt et al. 1999). This view has been challenged by current morphophonetic research. A morphophonetic approach to the acoustic representation of de may add to the scientific discussion about the nature and role of prepositional constituents in French.
In what follows in this article, Section 1 summarizes the scientific debate about the French preposition de in current research. Section 2 provides an overview of recent and relevant research on morphophonetics and introduces current research on the processing of multi-word units, complex words and n-grams. Section 3 presents the phonetic realization and morphological functioning of de, while Section 4 presents the eye-tracking reading study and discusses the results with regard to current theoretical approaches to morphological processing. Section 5 presents the methods and procedure of the experimental reading task as well as the results of the statistical analysis. The results of both studies are compared and discussed in the last section: the discussion and conclusion.
2 The preposition de
Prepositions in Romance languages still comprise a heterogeneous and not clearly defined word class. They belong to the invariable word classes, that is to say, that they cannot undergo inflectional or derivational processes. The most common prepositions in French are à ‘of/to’, de ‘of’, avec ‘with’, pour ‘for’ and en ‘in’ (Grévisse et al. 2018). Other common French prepositions are, for instance, après ‘after’, depuis ‘since’, sur ‘on’, sans ‘without’ and dans ‘in’ (Grévisse et al. 2018).
Prepositions may operate as a functional device and as a subordination marker between a head element and a complement and may express various relations between heads and complements. In French, prepositions may, for instance, mark a nominal compliment, as in un individu sans foi ‘an individual without a faith’, an adjectival complement, as in habile dans son art ‘clever in his skills’ or a verbal complement, as in Aude s’efforce de mieux faire ‘Aude makes an effort to do better’ (Grévisse et al. 2018: 11). Still, it is disputed whether prepositions carry an inherent lexical or merely a grammatical meaning. The scientific discussion on the status of French prepositions centers on two approaches that take opposite perspectives, more precisely the structuralist and the generative approach. While prepositions are considered to be mere grammatical elements without a semantic value in structuralist works on prepositions (e.g., Brøndal 1950; Pottier 1962), they are classified as a lexical category in generativist research (e.g., Jackendoff 1973). In the current article, I adopt the view that the preposition de is not semantically empty and may indeed carry semantic meaning patterns. It seems improbable to assume that prepositions do not carry any semantic value. The French preposition de may express a wide range of semantic meaning patterns that will be introduced in what follows.
In present-day French, the preposition de appears in various functions and expresses many different semantic relations. On the one hand, de functions as a free lexeme, for example as a partitive article in une tasse de thé ‘a cup of tea’ and un pot de confiture ‘a jam jar’ or as a free preposition in la réparation prendra de huit à dix jours ‘the repairs will take eight to ten days’ (Grévisse et al. 2018: 91–94). Furthermore, de as a free lexeme can introduce an infinitive in on permet de fumer ‘smoking permitted’, may function as an adjectival attribute marker in deux jours de libre ‘two days off’ or a possessive marker in le livre de mon père ‘my father’s book’ (Grévisse et al. 2018). The preposition de is also found in N Prep N constructions, as in pomme de terre ‘potato’ or in routinized phrases, such as de rien ‘you’re welcome’ (Grévisse et al. 2018). On the other hand, de also appears as a bound constituent in grammaticalized or lexicalized constructions: in complex prepositions (locutions prépositives), such as afin de ‘in order to’, avant de ‘before’ or près de ‘near’, and adverbial constructions (locutions adverbiales), such as du fait de ‘due to’ or de loin ‘far’, de is a bound constituent in grammatical or lexical constructions. In N Prep N constructions, such as salle de bain ‘bathroom’ or pomme de terre ‘potato’, de actually functions as a bound constituent in a clearly lexicalized construction (see Grévisse et al. 2018). It has to be mentioned that the general status of compounds and the classification of N Prep N constructions in compounding is a difficult issue in Romance linguistics. N Prep N constructions appear frequently in Romance languages as complex word forms referring to a specific concept, such as French pomme de terre ‘potato’, and as nominal phrases, such as French robe pour femmes ‘dress for women’. Since both types of constructions share the same overall structure, the delimitation and classification of these items is far from clear (for a discussion, see e.g., Hennecke 2019).
The present article aims at examining whether the different functions and positions of the clitic de show differences in their visual recognition and their phonetic realization in a reading task and in their acoustic realization. Here, the focus will be on the question if de shows specific acoustic peculiarities that may be linked to the morphosyntactic properties of the whole constructions. In that sense, this article contrasts the acoustic realization of de in free constructions with de in lexicalized and grammaticalized constructions. As mentioned earlier, four different functions of de will be compared in the following analyses. First, free syntactic occurrences of de, as described in the previous paragraph, represent de as a free lexeme in a syntactic structure. Second, de as a constituent in N Prep N compounds represents de as a morphemic element in a complex lexicalized construction that, as a whole, functions as a lexemic unit. Third, de in complex prepositions represents de in grammaticalized or lexicalized complex constructions. Complex prepositions, or locutions prépositives, are very frequent in French and other Romance languages and generally consist of complex constructions that function as prepositions and include at least one prepositional constituent, as in en face de ‘in front of’ or au travers de ‘through’. Fourth, in the same way, de in adverbial constructions, so-called locutions adverbiales, presents a constituent in grammaticalized or lexicalized constructions, as in moins de ‘less’ or tout de suite ‘immediately’.
The scientific discussion about the lexicalization or grammaticalization of prepositional and adverbial constructions still remains inconclusive precisely because these constructions occupy a position at the interface of the lexicon and syntax (see e.g., Gross 2006; Melis 2003 for a discussion of locutions prépositives in French and Antoni 2017 for a discussion on locutions adverbiales in French). Therefore, an experimental analysis of these constructions may potentially be informative. According to Fagard (2006: 44), adverbs, such as devant ‘in front of’ and prepositional constructions, such as au-devant de ‘in front of, ahead of’ are less lexicalized and more grammaticalized than prepositional syntagms, such as dans la direction de ‘in the direction of’. In that sense, Fagard and de Mulder (2007) propose that the emergence of prepositional constructions in French includes processes both of lexicalization and grammaticalization. Furthermore, Fagard (2006) mentions that different prepositional constructions in French may show different degrees of grammaticalization or lexicalization. For instance, according to him, à travers de ‘through’ is more grammaticalized than au travers de ‘through’ (Fagard 2006: 96). Some prepositional constructions allow internal variation and modification, such as à côté de, aux côtes de, du côté de ‘next to’ or constructions containing French bout ‘bit’, such as en bout de, (tout) au bout de ‘at the end of’ or à l’autre bout de ‘at the other end of’ (Borillo 1997). Other prepositional constructions allow insertion, such as outre que ‘other than’ and outre le fait que ‘other the fact than’ (Borillo 1997).
The same holds true for the group of adverbial constructions. Some adverbial constructions, such as tout de suite ‘now, right away’ can be clearly classified as lexicalized items that underwent the traditional processes of lexicalization and now function as a lexical elements in Modern French. Other adverbial constructions occupy a position on the lexicon-grammar continuum that is closer to the grammar pole of the cline. Adverbial constructions that function as quantifiers, such as beaucoup de ‘a lot of, peu de ‘little of’ or plein de ‘a lot of’, also take grammatical functions and therefore cannot be categorized as mere lexical items. For the reasons mentioned above, the present article assigns an intermediate position to complex prepositions and adverbial phrases as they cannot be unequivocally classified as lexical or grammatical word class. This is mainly due to the fact that the class of adverbial and prepositional phrases includes a wide range of heterogeneous elements that underwent and undergo different stages and processes of lexicalization and grammaticalization. Even though the exact degree of grammaticalization or lexicalization of complex prepositions and adverbial constructions remains unclear, a first explorative experimental investigation of these constructions in comparison with more lexicalized constructions, such as N Prep N, and free syntactic occurrences of de, is potentially informative.
3 Processing of constituents in multiword units
In the last decades, psycholinguistic research has tried to determine the specific nature of lexical processing of complex words and multi-word units. One of the main issues in this research is whether readers perceive complex words as a whole or if they are automatically decomposed into constituents during language processing. Decompositional accounts opt for early morphological decomposition and assume that, in language processing, the frequency effects of the individual constituents take precedence over the frequency effects of the whole complex word (e.g., Rastle et al. 2004; Taft and Forster 1976; see Rastle and Davis 2008 for a review). Some of these studies argue for task-dependent effects that only occur in lexical decision tasks (Duñabeitia et al. 2011). Other research suggests that the whole complex word is perceived first and that whole-word effects precede constituent effects (Giraudo and Grainger 2001). Yet other studies propose a parallel activation of whole-word processing and decompositional processes (Frauenfelder and Schreuder 1992; Schreuder and Baayen 1995). These studies suggest that two access routes operate independently and hence function as a dual route model. Early studies on this issue have been complemented by newer research with eye-tracking studies. The advantage of eye-tracking studies is that they allow a retracing of the exact temporal sequencing of the gaze behavior of the readers. As gaze movement is assumed to be closely linked to visual language processing, eye-tracking studies provide a more precise picture of the processing of constituents and multi-word units. Early eye-tracking studies present evidence that both constituents and whole words are relevant in visual word recognition (Hyönä et al. 2004; Zwitserlood 1994), which provides further evidence for dual-route models. More recent studies critically examine the classical stage models of morphological processing, in which the different frequency effects are interpreted as evidence for a specific stage in language processing (Baayen et al. 2009). Baayen et al. (2009) argue in favor of a more complex dynamic system of morphological processing that considers multiple dynamically interacting factors. In this vein, Baayen et al. (2007) interpret the full-form frequency effects in compounds as an indication of the reader’s knowledge about the probability of co-occurrence of the head constituent and the modifier constituent of a complex word. Schmidtke et al. (2017) find further evidence for whole-word effects preceding constituent effects in temporal sequencing in eye-tracking and lexical decision data.
All of these above-mentioned studies focus on complex words, mostly on NN compounds that can be semantically divided into a head constituent and a modifier constituent. However, it is of great interest to investigate the processing of less prototypical word classes as constituents in complex words.
A recent line of psycholinguistic research moves the focus away from clearly lexicalized complex words to multi-word units and n-grams. Various recent studies have found evidence for frequency effects in word n-grams (Arnon and Snider 2010; Bannard and Matthews 2008; Shaoul et al. 2013). Arnon and Snider (2010) found frequency effects for compositional four-word phrases, such as don’t have to worry, in that more frequent phrases are processed faster. In that sense, readers are sensitive to information on multi-word units, such as the probability of co-occurrence of lexical items and the productivity of specific multi-word schemas. The present study aims to add to this recent line of research by investigating the French preposition de in free syntactic constructions and as a constituent in different constructions.
Theoretical approaches assuming whole-word storage of lexicalized and grammaticalized constructions also expect a general difference in the processing of lexical and syntactic constructions. Traditional dual-system theories act on the assumption that non-compositional complex words are processed differently than compositional forms. To this effect, Ullman (2001) assumes that lexicalized forms that are stored directly in the lexicon are learned differently than grammatically computed forms and even rely on different neural processes. In contrast, single-system theories do not distinguish between lexical (“stored”) and syntactic (“computed”) constructions (for an overview see Snider and Arnon 2012: 128). Snider and Arnon (2012) question the clear distinction between stored and computed forms and state that it is not always obvious how to distinguish between these two types. Moreover, newer computational models do not even rely on the lexicon as a single system anymore (e.g., Baayen et al. 2019).
For the purpose of the present study, it is, therefore, crucial to establish several working hypotheses about the possible outcomes of the study. On the one hand, the present study aims at analyzing the role of the preposition de in language processing. As mentioned before, de has often been classified as an empty or colorless preposition, lacking any kind of semantic content and only functioning as a mere linking element. If de functioned as a semantically empty linking element, then it should not often be fixated in the course of reading. This should be particularly true for de as a constituent in complex words and multi-word units. In contrast, I assume that, if de is not a mere ‘empty’ preposition but a very polysemous unit, then de should be fixated equally in all construction types. On the other hand, the present study also looks at the processing of lexical items as constituents in multi-word units and in free syntactic constructions. Here, the focus will be on the question of whether de as a constituent in lexicalized or grammaticalized constructions is perceived differently than as a lexical item in free syntactic constructions. As mentioned above, there is no general consensus on the question of if and how lexicalized items, that are supposed to be stored in the mental lexicon, are processed differently than syntactic structures, which have to be combined through grammatical rules. The present study aims at investigating this issue on the constituent level. It is assumed that if there is a general processing difference between syntactic and lexicalized/grammaticalized constructions, then there should be visible processing differences in the constituent de. If the constituent de is processed similarly in all the different construction types, then this result may give additional support for theoretical approaches that do not assume a processing difference between stored and computed forms. The above-mentioned assumptions will be evaluated in what follows.
4 Phonetic realization and morphological functioning
The scientific discipline of morphophonetics and morphophonology investigates the interrelation of the morphosyntactic function of a lexical or morphological unit and its phonetic realization. Previous phonetic studies on lexemes and morphemes have shown that the function of a morpheme or a lexeme may indeed change its acoustic realization.
On a lexical level, the study by Drager (2011) analyzed the phonetic realization of English like as a quotative, as a discourse marker, and as a verb in utterances by female speakers from a high school in New Zealand. On the one hand, she investigated the relationship between the realization of specific phonetic variables and the social group of the individual speaker. On the other, she analyzed the interrelation between the acoustic realization of like and the individual speaker’s probability of using like for a specific function. The results show that the phonetic realization of like varies according to its grammatical function as well as the social group to which the individual speaker belongs. Drager (2011: 706) points out that the results of her study are challenging for theoretical models of speech production, such as Levelt et al. (1999), that give the same phonetic representation for homophonous and polysemous words. In one of the only morphophonetic studies on constructions in French, Pekarek Doehler (2016) studied the French construction je sais pas as a construction involving an epistemic claim. She investigated the construction je ne sais pas in audio recordings of an informal focus group discussion and coded the occurrences according to syntactic, morphophonetic, and discourse functional criteria. Morphophonetic variations of je ne sais pas include forms like je sais pas, j’sais pas, chais pas and ch’pas. The results document an interconnection between the acoustic reduction of the construction, its syntactic and morphophonological properties, its epistemic meaning, its placement and its interactional functioning. The studies from Drager (2011) and Pekarek Doehler (2016) account for the fact that there are systematic differences in the acoustic realization of polysemous lexemes as well as of morphophonetic variations of constructions and routinized formulas.
Blazej and Cohen-Goldberg (2015), Kemps et al. (2005), and Plag et al. (2017) conducted morphophonetic studies on morphemic elements, proving that the internal structure of words influences their articulation and acoustic realization. Plag et al. (2017) investigated the realization of the sounds/s/and/z/in morphemic and non-morphemic functions (e.g., as a clitic, genitive marker or plural marker vs. as a part of a lexical word). In their study based on data from the Buckeye Corpus, they include several covariates, such as speech rate, voicing, neighborhood density and various frequency measures in order to rule out any kind of disturbing influence on the acoustic realization of/s/and/z/. The results of the study indicate that the realization of/s/and/z/varies between non-morphemic and morphemic sounds as well as also between morphemes.
Another set of studies revealed that the syntagmatic and paradigmatic structure of morphological units may influence the phonetic realization of complex words (see, for instance, Bell et al. 2021; Kuperman et al. 2007; Plag and Balling 2017; Smith et al. 2012). Here, the notions of syntagmatic and paradigmatic structure are employed in a Saussurian sense, where syntagmatic relations occur between linearly arranged items in an utterance, and paradigmatic relations are created between interchangeable elements in the linguistic system as such. This set of studies generally focuses on the morphological boundary in inflectional and derivational affixes (e.g., Smith et al. 2012) or on complex word structures, such as compounds (Bell et al. 2021; Kuperman et al. 2007).
It can be stated as an interim conclusion that many of the studies in the last decade focus on the interconnection of phonetic detail, such as duration and stress, and morphological structure, more precisely derivational and inflectional affixes and complex words, such as compounds. In their fundamental theories and results, they challenge established models of speech processing, particularly of speech production (e.g., Levelt et al. 1999). For quite some time, there was a broad consensus in linguistic research on speech processing that the process of lexical and morphosyntactic planning had to be clearly distinguished from the post-lexical process of articulation. In this sense, it was unimaginable to suggest that morphological information could have any influence on the phonetic realization of a morphological or lexical item. As stated above, a large set of studies from the last decade challenge this well-established view and show numerous results from empirical and experimental studies that underline a new understanding of the interconnection of morphological information and phonetic detail.
In what follows, Sections 4 and 5 present the methods, data, and results of the experimental investigation of this study.
5 Eye-tracking study
5.1 Materials
The data for the eye-tracking study with a reading task were taken from the corpus Phonologie du français contemporain ‘Phonology of contemporary French’, PFC.[1] This corpus contains 49 surveys of French speakers from different geographic areas of the francophone world. For the purpose of the present study, only data from France were considered. Sentences containing de were extracted from the subcorpus of free conversations (conversations libres). Four construction types including de were selected for analysis, and 125 utterances containing de were extracted from the corpus data. These 125 utterances were classified according to the four construction types mentioned earlier.
| Example | Translation | Type | |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
The corpus of de constructions for this study was balanced as follows: 25 adverbial construction sentences, 25 prepositional construction phrases, 25 N Prep N compounds, and 50 phrases for de as a free preposition. All sentences were selected according to their readability, coherence, and internal structure (hesitation and reparation phenomena, markers of oral speech, fluidity of speech, etc.). Some sentences had to be slightly modified to fulfill the criteria of a readable sentence. The modifications never affected the prepositional element de or its direct environment. Additionally, 200 filler sentences were extracted from the PFC corpus that did not include any occurrence of the preposition de.
Furthermore, the data were classified according to the boundedness of the preposition de in the respective construction. In example c., de clearly functions as a free lexical unit in a phrase. In contrast, in examples a., b. and d., de is a bound constituent in a lexicalized/grammaticalized construction. Therefore, de as a free preposition was coded as a free constituent in a phrasal structure, whereas de in prepositional, adverbial, and compound structures was coded as a bound constituent.
5.2 Participants
A total of 18 native speakers of European French participated in the task in exchange for payment. Before the task, all participants signed a consent form that informed them about the procedure of the task. All participants reported having normal vision.
5.3 Procedure
The eye-tracking task was conducted using an EyeLink 1000 tracking device.[2] The recording of the eye movements was performed on the right eye only. Participants’ calibration of the pupil took place at the beginning of the experiment, with recalibration after each block of 25 stimuli.
Each trial started with the presentation of a fixation cross at the left part of the center of the screen. The participant had to fixate the cross in order to start the trial. Then the stimulus sentence was presented starting at the left part of the center of the computer screen. The participants were asked to read every sentence out loud in order to ensure the reading of the sentence and to record the audio for the reading study. To start a new trial, the participants were asked to fixate on a white circle at the right-center of the screen after reading the stimulus sentence. This fixation or trial duration of 3 s triggered the start of the next trial.
5.4 Predictor variables
The analysis measured fixation duration and fixation counts of the constituent de in different lexical environments. Eye movements consist of fixations and saccades. During fixations, the eye rests on the visual input and the information processing is secured. The fixation duration measures the time the eye rests on the visual input during one specific fixation. During saccades, the eye moves fast from one point to another. The fixation count measures the number of fixations on a certain lexical item during reading. The present analysis includes the first and second fixation durations on de as well as the fixation counts in the first fixation run and in the second fixation run. According to the official Data Viewer Manual from SR Research (2019: 20), a fixation run is defined as “two consecutive fixations in the same interest area”. In the same way, the fixation count is defined as the “total number of fixations within the interest area (instance) in the current interest period” (SR Research Ltd 2019: 71). For the present analysis, the interest period was set as the period within a trial before regressive eye-movements for refixations (see SR Research Ltd 2019: 3).
All measures of fixation duration and fixation counts were obtained from the Interest Area Report that includes the respective information for each Interest Area. Each constituent of the sentence was assigned an Interest Area, and all fixation points that lay outside of an Interest Area were excluded from the analysis.
Furthermore, the analysis included various frequency variables, specifically of the construction, e.g., beaucoup de ‘a lot’; the frequency of the whole expression, e.g., beaucoup de gens ‘a lot of people’; the frequency of the word preceding de; and the frequency of the word following de.
In addition, the analysis coded for four factorial predictors: the lexical status of the construction was coded as a factor with the two levels grammatical and lexical. The boundedness of a morpheme was coded as a factor with the two levels bound and free. Furthermore, the type of the construction was coded as a factor with four levels, viz. adverbial construction, complex prepositions, N Prep N constructions, or free preposition.
All analyses included random intercepts that control for variance within a specific factor for both the participant and the stimulus. All analyses included random intercepts that control for variance within a specific factor for both the participant and the stimulus.
5.5 Results
Statistical analyses of the data were performed using R for statistical computing (R Core Team 2017). Statistical analyses were only performed for the experimental stimuli data, using generalized additive models as implemented in the mgcv package (Li and Wood 2020). Generalized additive models were used for all analyses of the fixation durations and analyses of the fixation counts. As the analysis of the fixation durations revealed no significant effects, only the analysis results of the fixation counts are reported here. The lack of significant effects in the analysis of the fixation durations may be due to the monomorphemic nature of de.
5.5.1 Non-fixations
The dataset includes a total of 2470 appearances of de. The distribution of the appearances is visualized in the following table (Table 1):
In the first fixation run, de constructions were fixated in a total of 1087 cases (44%), i.e., non-fixated in a total of 56% of cases. The distribution of the omissions in the first fixation run is shown in Table 2:
Overall occurrences of de in the stimuli sentences.
| Construction | N |
|---|---|
| Adverbial construction | 494 |
| Compounds | 494 |
| Free prepositions | 969 |
| Complex prepositions | 513 |
| Total | 2470 |
-
The bold values are the total values of all constructions mentioned above (Total) in absolut numbers (N).
Non-fixations of de in the first fixation run of the stimuli sentences.
| Construction | N | % |
|---|---|---|
| Adverbial construction | 257 | 52 |
| Compounds | 259 | 52.4 |
| Free prepositions | 507 | 52.3 |
| Complex prepositions | 324 | 63 |
| Total | 1383 | 100 |
-
The bold values are the total values of all constructions mentioned above (Total) in absolut numbers (N) and percentage (%).
This table shows that there is a relatively even distribution of non-fixations across the different construction types in which de appears. The preposition de is most frequently omitted in complex prepositions and least frequently in adverbial construction. However, a post-hoc test revealed no significant difference between these two construction types.
The second fixation run gives a different picture. Out of all the 2470 occurrences of de, only 243 instances (10%) are fixated twice, meaning that in 2227 cases (90%) there is no second fixation. Table 3 shows the distribution of the non-fixations of de across construction types.
Non-fixations of de in the second fixation run of the stimuli sentences.
| Construction | N | % |
|---|---|---|
| Adverbial construction | 433 | 87.6 |
| Compounds | 456 | 92.3 |
| Free prepositions | 859 | 88.6 |
| Complex prepositions | 479 | 93.3 |
| Total | 2227 | 100 |
-
The bold values are the total values of all constructions mentioned above (Total) in absolut numbers (N) and percentage (%).
Table 3 again shows a relatively even distribution. While just under half of the de constructions are fixated in the first run, only some 7–13% of the cases are fixated in the second run. In conclusion, the investigation of non-fixations of the preposition de does not indicate any significant differences in the omission of the preposition de. That is to say that the data show an even distribution of non-fixations across the construction types and do not suggest any processing differences between construction types.
5.5.2 Analysis of the fixation counts
For the analysis of the fixation counts, the analysis is based on generalized additive mixed models (GAMM) with a Poisson link to model the counts of fixations on the preposition. In the analyses, predictors that were not well supported were removed from the models. The model summaries presented here are of these trimmed models. All analyses are part of an explorative inquiry into the structure of the experimental data.
The following two tables represent the results from the Poisson generalized additive models fitted to the fixation counts for the preposition. Table 4 shows the results from the analysis of the first fixation counts, while Table 5 presents the results from the analysis of the second fixation counts.
Analysis of first fixation counts.
| Parametric coefficients: | Estimate | Standard. Error | t-value | Pr(>|t|) | |
|---|---|---|---|---|---|
| (Intercept) | 34.219 | 0.2677 | 12.781 | <2e-16 | *** |
| N Prep N | 10.743 | 0.3705 | 2.900 | 0.00382 | ** |
| Free preposition | 0.4353 | 0.3207 | 1.357 | 0.17496 | |
| Complex prepositions | 10.953 | 0.3712 | 2.951 | 0.00325 | ** |
| Smooth terms: | edf | Ref.df | F | p-value | |
|---|---|---|---|---|---|
| s(Wholefreq) | 5,16E-03 | 1 | 0.000 | 0.203172 | |
| s(Constrfreq) | 2,58E-02 | 1 | 0.000 | 0.419754 | |
| s(Word) | 1,07E+05 | 120 | 9.053 | <2e-16 | *** |
| s(Subject) | 1,04E+04 | 18 | 1.567 | 0.000335 | *** |
Analysis of second fixation counts.
| Parametric coefficients: | Estimate | Std. Error | t value | Pr(>|t|) | |
|---|---|---|---|---|---|
| (Intercept) | 117.53 | 0.1 | 13.0 | <2e-16 | *** |
| N Prep N | −0.12 | 0.1 | −2.0 | 0.045 | * |
| Free preposition | −0.06 | 0.0 | −1.4 | 0.146 | |
| Complex preposition | −0.02 | 0.1 | −0.3 | 0.701 |
| Smooth terms: | edf | Ref.df | F | p-value | |
|---|---|---|---|---|---|
| s(Wholefreq) | 3,3 | 1 | 0.0 | 0.742 | |
| s(Constrfreq) | 5,9 | 1 | 1.5 | 0.115 | |
| s(Word) | 4,7 | 82 | 0.0 | 0.726 | |
| s(Subject) | 1,6 | 18 | 3.5 | 1.17e-06 | *** |
Distribution of types of de in the experimental reading task data.
| Type | N |
|---|---|
| Adverbial construction | 312 |
| N Prep N | 289 |
| Free preposition | 447 |
| Complex prepositions | 279 |
| Total | 1327 |
-
The bold values are the total values of all constructions mentioned above (Total) in absolut numbers (N).
As mentioned earlier, the fixation count is defined as the “total number of fixations within the interest area (instance) in the current interest period” (SR Research Ltd 2019: 71). For the present study, the interest period was set as the period within a trial before regressive eye-movements for refixations (see SR Research Ltd 2019: 3). It is important to recall that only about half of the prepositions were fixated in the first fixation counts, whereas the other half was not fixated at all during reading. Therefore, the analysis of the fixation counts is only of the prepositions that were fixated at least once. The analysis of the first fixation counts of the preposition de shows significantly higher fixation counts for de in N Prep N constructions and complex prepositions than in adverbial constructions. Free prepositions do not show a significant effect. That is to say, if de is fixated in the first fixation counts, then it is fixated more often in N Prep N constructions and complex prepositions than in adverbial phrases and as a free preposition. Furthermore, the analysis shows no frequency effects nor any significant effects of the predictor variables lexical status and boundedness. In other words, the fixation counts do not show any significant difference in the processing of the preposition de as an item in free syntactic or lexicalized environments. Moreover, the data show no processing difference of de as a bound constituent in stored multi-word units or as a free constituent in a syntactic structure.
The analysis of the second fixation counts shows a somewhat different picture:
Again, it is important to remember that only 6.7% of de in prepositional constructions, 7.7% of de in compounds, 11.4% of de as a free preposition, and 12.4% of de in adverbial constructions were fixated twice. The analysis of the second fixation counts, which is clearly based on a small dataset, reveals a reversed effect from the first fixation counts. The second fixation counts also show a significant effect for de in compounds, but this time, the effect goes in the opposite direction and shows significantly fewer fixation counts for de in compounds than in adverbial constructions. Here again, the analysis does not reveal any significant frequency effects and does not show any effects for the other predictor variables lexical status and boundedness.
5.5.3 Discussion
In conclusion, we see that both analyses show a significant effect of the construction type in which de occurs. The first interesting result of the study is that only the fixation counts show significant results. The data show that the fixation counts are more meaningful for the analysis of de than the fixation durations. The lack of effects in the fixation duration may be due to the monomorphemic nature of de.
Furthermore, the study reveals that about half of the occurrences of de are fixated in the first fixation count. Here, the data show an even distribution among the different construction types. The second fixation count only shows a few fixations on the preposition, which are again distributed evenly across the different construction types. This result underlines that the preposition de is only fixated in approximately half of the first fixations and if it is fixated, then mostly only once. This result may be interpreted with regard to different theoretic approaches. As de is fixated in roughly half of the cases, it can often be interpreted in the way that de is not overlooked while reading and hence carries important information for language processing. From another perspective, the result may be interpreted as the exact opposite. As de is only fixated in approximately half of the cases, this result may be interpreted as evidence for the fact that the fixation of de is not crucial for reading. However, as de is a monomorphemic unit, it is conceivable that in some cases, de is already fixated with a parafoveal view while reading the preceding constituent. In the end, the data do not give enough evidence to interpret this finding unequivocally.
However, it is quite striking that the data do not show any processing differences for the different construction types with regard to the non-fixations on the preposition. That is to say, none of the construction types trigger any processing differences. As mentioned in Section 2, similar processing of the constituent de in the different construction types might provide new evidence for theoretical approaches that do not assume a processing difference between stored and computed forms. In other words, this result argues against a clear distinction between stored (lexicalized) and computed (syntactic) constructions and also against traditional dual-system approaches to language processing.
The statistical analysis of the fixation counts reveals further interesting results. As already mentioned, the data for the underlying analysis varied in size between the first and second fixation counts. Moreover, the analysis of the first fixation counts revealed a significantly higher number of fixation counts for de in N Prep N constructions and complex prepositions than in adverbial constructions. This result is quite surprising, as N Prep N constructions as clearly lexicalized constructions show the highest fixation counts on the constituent de. This result may argue against traditional “whole storage”’ approaches and dual-system approaches, in which complex words are assumed to be stored and accessed as a whole in language processing. However, the result changes for the analysis of the smaller data set of the second fixation count. The analysis of the second fixation count shows that if de is re-fixated in a second run, then the least fixations fall on de in N Prep N constructions, compared to adverbial constructions. One possible explanation for this result is that de in N Prep N constructions and in complex prepositions shows a clear relational functioning in that it expresses the semantic relation between the previous and the following lexical element, for instance between the two nominal constituents in the N Prep N construction. As de is a very polyfunctional and polysemous unit, it may express various kinds of relations, such as a target group, as in livre d’enfant ‘children’s book’ (N1 is made for N2), a part-whole relation, as in collier de perles ‘pearl necklace’ (N2 is a part of N1) or a location, as in route de montagne ‘mountain road’ (N1 is located in N2) (see Hennecke 2019 for a more detailed and comparative qualitative corpus analysis). However, more research on the role of prepositional constituents in multi-word units is necessary in order to unequivocally explain the peculiarities in the processing of de in N Prep N constructions and complex prepositions in the first fixation counts.
Another interesting result of the present study is that no frequency effects appear in the gaze durations and fixation counts of the prepositional constituent de. The lack of frequency effects may be due to the monomorphemic structure of de. Nevertheless, more research on the role of prepositional constituents in multi-word units is needed in order to unequivocally clarify this issue.
6 Experimental reading task
6.1 Data
The present phonetic investigation is based on the same data as the eye-tracking study in the previous section.
6.2 Procedure
A total of 17 native speakers of European French participated in the task in exchange for payment. All participants were unaware of the aim of the task and signed a consent form that informed them about the procedure and the risks of the task. All participants were seated in a soundproof booth in front of a computer screen and a microphone. Stimuli sentences were presented on the computer screen in light grey writing on a black background. Each trial started with the presentation of a fixation cross and the appearance of the stimulus sentence. The participants read every stimulus out loud and their speech was recorded.
6.3 Analysis
For the analysis, the acoustic duration was considered the dependent variable. The acoustic duration of the individual lexemes of de was obtained through automatic segmentation of the data with the Praat (Boersma and Weenink 2019) plugin Easyalign and manual verification of the segmentation (Eychenne 2018).[3] Filler sentences and sentences containing nonfluent speech, hesitations, or repairs reparations were excluded from the analysis. This resulted in a total of 1327 occurrences of de that qualified for analysis. The following table represents the distribution of the types of de in the data set (Table 6):
The distribution and means of the duration (in ms) of de by type are given in Figure 1:
Duration (in ms) of the preposition de by type. Lines represent the mean of the type.
For the present analysis, the same sets of acoustic and non-acoustic covariates as in the analysis of the natural speech data were included. This includes various frequency measures, more precisely the frequency of the whole unit (e.g., beaucoup de neige ‘a lot of snow’), the frequency of the construction (e.g., beaucoup de ‘a lot of’), as well as the frequencies of the word preceding and the following de. All frequency measures were log-transformed for the statistical analysis in order to secure a homogeneous distribution. The number of letters in the words neighboring de, as well as the position of de in the construction, were also included as covariates. The last covariate is the speech rate of the individual speaker. The speech rate was again calculated by counting the number of syllables of the whole utterance of the speaker and by dividing this number by the length of the utterance (in seconds).
6.4 Statistical analysis
The statistical analysis of the data was performed using R software for statistical computing (R Core Team 2017). Statistical analyses were performed on the text grids containing the phonetic duration of the preposition de. The results were analyzed using generalized additive models in the mgcv package in R (for further information on the statistical framework see Li and Wood 2020). Generalized additive models were chosen because of the possibility of non-linearity of the data.
6.5 Results
Table 7 summarizes the results of the statistical analysis with generalized additive models. The results indicate significant frequency effects for the construction frequency, the frequency of the whole construction as well as for the frequency of the word preceding de. The frequency effect is a well-known and extremely well-studied effect in visual word recognition and reading. The results in Table 7 show that de in more frequent constructions is processed faster than de in less frequent constructions. The same effect holds true for the frequency of the whole expression. Furthermore, the frequency of the word that precedes de shows a significant effect in that a higher frequency of the previous word leads to slower articulations of de.
Summary of the generalized additive model of the acoustic durations of the preposition de.
| Parametric coefficients | Estimate Std. | Error | t-value | Pr(>|t|) |
|---|---|---|---|---|
| (Intercept) | 5.558e+00 | 9.057e-02 | 61.371 | <2e-16*** |
| Construction frequency | −1.252e-02 | 3.407e-03 | −3.674 | 0.0002*** |
| Whole frequency | −6.570e-03 | 2.269e-03 | −2.896 | 0.0038** |
| Frequency previous word | 1.660e-09 | 3.808e-10 | 4.360 | 1.4e-05*** |
| N Prep N | −9.792e-02 | 3.302e-02 | −2.966 | 0.0031** |
| Free preposition | −1.565e-03 | 2.484e-02 | −0.063 | 0.9497 |
| Complex prepositions | −4.045e-02 | 2.502e-02 | −1.617 | 0.1061 |
|
|
||||
| B. Smooth terms | edf | Ref.df | F | p-value |
|
|
||||
| s (speech rate) | 0.9901 | 1 | 99.52 | <2e-16*** |
We find a significant effect for de as a constituent in N Prep N constructions. In the present study, de in N Prep N constructions is articulated significantly faster than in adverbial constructions. This result holds true in a post-hoc test in the form of a Tukey test. Figure 2 presents the results:
Results from the post-hoc Tukey test.
Figure 2 shows that only the difference of realization of de between compounds and adverbial phrases is significant, in that de in compounds is articulated significantly faster than in adverbial constructions. The other tests in the Tukey test do not show significant results.
Further, the present study does not show any effect on the boundedness of de or its lexicality. Similarly, none of the other covariates mentioned in the previous part of the article show any significant effects.
6.6 Discussion
This experimental study aimed at investigating the morphophonetic realization of the French clitic de in different morphemic and lexical environments. Previous studies have shown that the acoustic realization of lexemes and morphemes may be influenced by their morphosyntactic functioning and their morphemic status. The results of the present study provide further support for these previous findings, in that de in lexicalized N Prep N constructions behaves differently than in the other observed environments of de.
The results from the reading study reveal various frequency effects for the construction and the whole expression in which de is embedded. These frequency effects show that the acoustic realization of de in a reading task is influenced by the frequency of the whole construction as well as by the frequency of the word that precedes de. Moreover, the analysis of the data from the reading task also reveals a significant effect on the acoustic realization of de in N Prep N constructions, in that de in these constructions is processed significantly faster than in adverbial constructions.
The analyses also show the importance of controlling for various types of covariates, such as the individual speaker, the speech rate, and the phonetic environment of the linguistic unit under investigation. As expected, the individual speaker and the speech rate of the speaker may have a significant effect on the phonetic realization of the linguistic unit de. Furthermore, the acoustic duration appears to be influenced by the nature of the processing of linguistic information.
7 Conclusion
In conclusion, the present study reveals several interesting insights into the processing of the constituent de in different kinds of multi-word units and syntactic environments. As this study is explorative in nature, it can only present tendencies that must be analyzed in more detail in future research. These tendencies are summarized here.
In the eye-tracking study, the preposition de is fixated in about half of the cases with an equal distribution across construction types. This result cannot be interpreted unequivocally and can be seen both as evidence for and against approaches that argue against “colorless” prepositions without semantic content that function as mere linking elements. Further, the eye-tracking study reveals no processing difference between lexicalized/grammaticalized and free constructions nor between de as a bound or a free morpheme. This result is interpreted here in favor of approaches arguing against dual-system theories, more precisely against a clear processing difference of ‘stored’ and ‘computed’ forms. However, the lack of a significant effect may also be the result of the fact that the degree of lexicalization or grammaticalization in many constructions cannot be clearly defined. Each type of construction in the analyses included more and less grammaticalized or lexicalized items, which may have influenced the outcome of the analyses. Moreover, the eye-tracking study revealed that de as a constituent in N Prep N constructions shows certain peculiarities in the analysis of the fixation count. This result is ambiguous, but it may be linked to the relational character of de in N Prep N constructions. In addition, no frequency effects are found in the analyses of the fixation durations and fixation counts of de. This result is probably linked to the monomorphemic structure of de. In order to investigate these issues in more detail, more research on the processing of prepositions is necessary. Future studies will need to investigate different prepositions in different types of multi-word units both from a monolingual but also from a comparative crosslinguistic perspective.
The phonetic analysis of the data set also revealed interesting results that may complement the results from the eye-tracking study. The phonetic analysis again shows certain peculiarities of the acoustic realization of de in N Prep N construction, in that it is articulated significantly faster than in adverbial constructions. In the same way as the eye-tracking study, the phonetic analysis did not show any significant differences in the realization of de in grammaticalized or lexicalized constructions. However, as mentioned above, the lack of effects may be due to the distribution of items in the data set. In addition, the phonetic analysis reveals the importance of different frequency effects that influence the articulation of de in specific constructions.
To conclude, this explorative study of de reveals emerging tendencies in the processing of prepositions in multi-word units and their acoustic realizations. First, the role of the variables influencing the processing of prepositional elements in multi-word units must be further investigated, particularly the role of lexicalization and grammaticalization. Second, the processing of prepositions as lexical or grammatical elements with or without a specific semantic content requires more detailed analysis. Third, the processing of prepositional constituents in N Prep N constructions appears to be particularly interesting and should be examined in depth in any upcoming investigations.
Supplementary Material
The dataset underlying this article is available from the Zenodo repository and may be accessed at https://doi.org/10.5281/zenodo.7115478.
References
Antoni, Marie-Hélène. 2017. Fonctionnement des locutions adverbiales du français: complexité interne et externe. In Cahiers Forell: Formes et Représentations en Linguistique et Littérature: Archives (1993–2001) Complexité syntaxique et sémantique. https://cahiersforell.edel.univ-poitiers.fr:443/cahiersforell/index.php?id=489 (accessed 20 September 2021).Search in Google Scholar
Arnon, Inbal & Neal Snider. 2010. More than words: Frequency effects for multi-word phrases. Journal of Memory and Language 62. 67–82. https://doi.org/10.1016/j.jml.2009.09.005.Search in Google Scholar
Baayen, R. Harald, Lee H. Wurm & Joanna Aycock. 2007. Lexical dynamics for low-frequency complex words. A regression study across tasks and modalities. The Mental Lexicon 2(3). 419–463. https://doi.org/10.1075/ml.2.3.06baa.Search in Google Scholar
Baayen, R. Harald, Victor Kuperman & Raymond Bertram. 2009. Frequency effects in compound processing. In Sergio Scalise & Irene Vogel (eds.), Cross-disciplinary issues in compounding, 257–270. Amsterdam & Philadelphia: John Benjamins.10.1075/cilt.311.20baaSearch in Google Scholar
Baayen, R. Harald, Yu-Ying Chuang, Elnaz Shafaei-Bajestan & James P. Blevins. 2019. The discriminative lexicon: A unified computational model for the lexicon and lexical processing in comprehension and production grounded not in (de)composition but in linear discriminative learning. In Complexity. https://www.hindawi.com/journals/complexity/2019/4895891/ (accessed 20 September 2021).10.1155/2019/4895891Search in Google Scholar
Bannard, Colin & Danielle Matthews. 2008. Stored word sequences in language learning: The effect of familiarity on children’s repetition of four-word combinations. Psychological Science 19(3). 241–248. https://doi.org/10.1111/j.1467-9280.2008.02075.x.Search in Google Scholar
Bell, Melanie J., Sonia Ben Hedia & Ingo Plag. 2021. How morphological structure affects phonetic realization in English compound nouns. Morphology 31. 87–120. https://doi.org/10.1007/s11525-020-09346-6.Search in Google Scholar
Blazej, Laura J. & Ariel M. Cohen-Goldberg. 2015. Can we hear morphological complexity before words are complex? Journal of Experimental Psychology: Human Perception and Performance 41(1). 50–68. https://doi.org/10.1037/a0038509.Search in Google Scholar
Boersma, Paul & David Weenink. 2019. Praat: Doing phonetics by computer [Computer program]. Version 6.0.46. http://www.praat.org/ (accessed 3 January 2019).Search in Google Scholar
Borillo, Andrée. 1997. Aide à l’identification des prépositions composées de temps et de lieu. Faits de langues 5(9). 175–184. https://doi.org/10.3406/flang.1997.1153.Search in Google Scholar
Brøndal, Viggo. 1950. Théorie des prépositions: introduction à une sémantique rationnelle. Copenhagen: Munksgaard.Search in Google Scholar
Drager, Katie. 2011. Sociophonetic variation and the lemma. Journal of Phonetics 39(4). 694–707. https://doi.org/10.1016/j.wocn.2011.08.005.Search in Google Scholar
Duñabeitia, Jon, Sachiko Kinoshita, Manuel Carreiras & Dennis Norris. 2011. Is morpho-orthographic decomposition purely orthographic? Evidence from masked priming in the same–different task. Language & Cognitive Processes 26(4–6). 509–529. https://doi.org/10.1080/01690965.2010.499215.Search in Google Scholar
Eychenne, Julien. 2018. Dolmen: A program for the analysis of speech corpora [Computer program]. Version 2.0.0. http://www.dolmen-ling.org (accessed 4 May 2018).Search in Google Scholar
Fagard, Benjamin. 2006. Évolution sémantique des prépositions dans les langues romanes: illustrations ou contre-exemples de la primauté du spatial? Paris & Rome: Université Paris 7 & Università Roma 3 dissertation.Search in Google Scholar
Fagard, Benjamin & Walter De Mulder. 2007. La formation des prépositions complexes: Grammaticalisation ou lexicalisation? Langue Française 4. 9–29. https://doi.org/10.3917/lf.156.0009.Search in Google Scholar
Frauenfelder, Ulrich & Robert Schreuder. 1992. Constraining psycholinguistic models of morphological processing and representation: The role of productivity. In Geert Booij & Jaap van Marle (eds.), Yearbook of morphology, 165–183. Dordrecht: Kluwer.10.1007/978-94-011-2516-1_10Search in Google Scholar
Giraudo, Hélène S. & Jonathan Grainger. 2001. Priming complex words: Evidence for supralexical representation of morphology. Psychonomic Bulletin & Review 8(1). 127–131. https://doi.org/10.3758/bf03196148.Search in Google Scholar
Grévisse, Maurice, Cédrick Fairon & Anne-Catherine Simon. 2018. Le petit bon usage de la langue française. Louvain: DeBoeck.Search in Google Scholar
Gross, Gaston. 2006. Sur le statut des locutions prépositives. Modeles Linguistiques 53. 35–50.10.4000/ml.517Search in Google Scholar
Hennecke, Inga. 2019. Internal constituent variability and semantic transparency in N Prep N constructions in Romance languages. In Sabine Schulte im Walde & Eva Smolka (eds.), The role of constituents in multiword expressions: An interdisciplinary, cross-lingual perspective, 131–156. Berlin: Language Science Press.Search in Google Scholar
Hyönä, Jukka, Raymond Bertram & Alexander Pollatsek. 2004. Are long compound words identified serially via their constituents? Evidence from an eye-movement-contingent display change study. Memory & Cognition 32. 523–532. https://doi.org/10.3758/bf03195844.Search in Google Scholar
Jackendoff, Ray. 1973. The base rules for prepositional phrases. In Stephen Anderson & Kiparsky Paul (eds.), A festschrift for Morris Halle, 345–356. New York: Holt, Rinehart & Winston.Search in Google Scholar
Kemps, Rachèl J. J. K., Mirjam Ernestus, Robert Schreuder & R. Harald Baayen. 2005. Prosodic cues for morphological complexity: The case of Dutch plural nouns. Memory & Cognition 33(3). 430–446. https://doi.org/10.3758/bf03193061.Search in Google Scholar
Kuperman, Victor, Mark Pluymaekers, Mirjam Ernestus & R. Harald Baayen. 2007. Morphological predictability and acoustic salience of interfixes in Dutch compounds. Journal of the Acoustical Society of America 121(4). 2261–2271. https://doi.org/10.1121/1.2537393.Search in Google Scholar
Levelt, Willem J. M., Ardi Roelofs & Antje S. Meyer. 1999. A theory of lexical access in speech production. Behavioral and Brain Sciences 22(1). 1–75. https://doi.org/10.1017/s0140525x99001776.Search in Google Scholar
Li, Zheyuan & Simon N. Wood. 2020. Faster model matrix crossproducts for large generalized linear models with discretized covariates. Statistics and Computing 30. 19–25. https://doi.org/10.1007/s11222-019-09864-2.Search in Google Scholar
Melis, Ludo. 2003. La préposition en français. Paris: Ophrys.Search in Google Scholar
Pekarek Doehler, Simona. 2016. More than an epistemic hedge: French je sais pas ‘I don’t know’ as a resource for the sequential organization of turns and actions. Journal of Pragmatics 106. 148–162. https://doi.org/10.1016/j.pragma.2016.06.014.Search in Google Scholar
Plag, Ingo & Laura W. Balling. 2017. Derivational morphology: An integrated perspective. In Vito Pirrelli, Wolfgang U. Dressler & Ingo Plag (eds.), Word knowledge and word usage: A cross-disciplinary guide to the mental lexicon, 296–335. Berlin & Boston: De Gruyter Mouton.Search in Google Scholar
Plag, Ingo, Julia Homann & Gero Kunter. 2017. Homophony and morphology: The acoustics of word-final S in English. Journal of Linguistics 53(1). 81–216. https://doi.org/10.1017/s0022226715000183.Search in Google Scholar
Pottier, Bernard. 1962. Systématique des éléments de relation. Paris: Klincksieck.Search in Google Scholar
Rastle, Kathleen, Matthew H. Davis & New Boris. 2004. The broth in my brother’s brothel: Morpho-orthographic segmentation in visual word recognition. Psychonomic Bulletin & Review 11(6). 1090–1098. https://doi.org/10.3758/bf03196742.Search in Google Scholar
Rastle, Kathleen & Matthew H. Davis. 2008. Morphological decomposition based on the analysis of orthography. Language & Cognitive Processes 23(7–8). 942–971. https://doi.org/10.1080/01690960802069730.Search in Google Scholar
R Core Team. 2017. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Available at: https://www.R-project.org/.Search in Google Scholar
Rumelhart, David E. & James L. McClelland & PDP Research Group. 1986. Parallel distributed processing. Cambridge, MA: MIT Press.10.7551/mitpress/5236.001.0001Search in Google Scholar
Schmidtke, Daniel, Kazunaga Matsuki & Victor Kuperman. 2017. Surviving blind decomposition: A distributional analysis of the time-course of complex word recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition 43(11). 1793–1820. https://doi.org/10.1037/xlm0000411.Search in Google Scholar
Schreuder, Robert & R. Harald Baayen. 1995. Modeling morphological processing. In Laurie B. Feldman (ed.), Morphological aspects of language processing, 131–154. Hillsdale, NJ: Lawrence Erlbaum.Search in Google Scholar
Shaoul, Cyrus, Chris F. Westbury & R. Harald Baayen. 2013. The subjective frequency of word n-grams. Psihologija 46(4). 497–537. https://doi.org/10.2298/psi1304497s.Search in Google Scholar
Smith, Rachel, Rachel Baker & Sarah Hawkins. 2012. Phonetic detail that distinguishes prefixed from pseudo-prefixed words. Journal of Phonetics 40(5). 689–705. https://doi.org/10.1016/j.wocn.2012.04.002.Search in Google Scholar
Snider, Neal & Inbal Arnon. 2012. A unified lexicon and grammar? Compositional and non-compositional phrases in the lexicon. In Dagmar Divjak & Stefan Th. Gries (eds.), Frequency effects in language representation, 127–164. Berlin & Boston: De Gruyter Mouton.10.1515/9783110274073.127Search in Google Scholar
SR Research Ltd. 2019. EyeLink data viewer user’s manual. Mississauga, Canada: SR Research Ltd.Search in Google Scholar
Taft, Marcus & Kenneth I. Forster. 1976. Lexical storage and retrieval of polymorphemic and polysyllabic words. Journal of Verbal Learning and Verbal Behavior 15(6). 607–620. https://doi.org/10.1016/0022-5371(76)90054-2.Search in Google Scholar
Ullman, Michael T. 2001. The declarative/procedural model of lexicon and grammar. Journal of Psycholinguistic Research 30(1). 37–69. https://doi.org/10.1023/a:1005204207369.10.1023/A:1005204207369Search in Google Scholar
Zwitserlood, Pienie. 1994. The role of semantic transparency in the processing and representation of Dutch compounds. Language & Cognitive Processes 9(3). 321–368. https://doi.org/10.1080/01690969408402123.Search in Google Scholar
© 2022 the author(s), published by De Gruyter, Berlin/Boston
This work is licensed under the Creative Commons Attribution 4.0 International License.
