Compounding is the most widespread word-formation strategy in languages worldwide, while English and German, as representatives of the Germanic languages, have a high preference for the use of compounds (Dressler, 2006). A compound is the result of a word-formation, where two or more words are combined (Hentschel, 2020). Besides compounding, a further word-formation strategy is the derivation, where a suffix is attached to a free morpheme (Schlücker, 2012) as seen in Les-er “read-er” or Les-ung “read-ing”. German compounding is highly productive, mostly leading to a combination of two or more words. According to Wellmann (1991), approximately 75% of the German noun vocabulary consists of compound-nouns. Some words are combined with linking elements of which the -s like in Wissenslücke “knowledge gap” is the most common in German (Hentschel, 2020). However, these elements will not be the focus of this paper. German compounds are mostly composed of two constituents and often appear without a linking element, as in Obstbaum “fruit tree” (Schlücker, 2012, p. 5). The right constituent, Baum “tree” in this example, functions as the head, determining the overall meaning of the compound. The first constituent, Obst “fruit”, provides additional specification and is referred to as the modifier. This example is known as an endocentric or determinative compound and accounts for 99.9% of German compounds (Hentschel, 2020). German and English endocentric compounds are realized right-headed, as illustrated in the example Obstbaum “fruit tree”. In other languages, like Romance languages, compounds are mostly realized left-headed. Contrary to endocentric compounds, there are exocentric compounds, which are unusual in German, but more common in languages like Italian and English (Schlücker, 2012). An exocentric compound noun does not have a clear head, because the word as a whole does not belong to the same word class or category as one of its components. The meaning goes beyond the individual components, as illustrates in Spaßbremse “killjoy” (Schlücker, 2012, p. 6). A further classification is the distinction between root and synthetic compounds. The head of a synthetic compound is always derived from a verb, like in Buchleser “book reader”, which describes the relation of meaning. For root compounds, like “fruit tree”, the semantic relation of the constituents is not obvious and has to be interpreted situationally (Müller et al., 2015). While this can cause challenges and misunderstandings due to unfamiliar words, it also enables the creation of new and unfamiliar terms through novel compounds, which is a key factor in the high productivity of German compounding. Additionally, the right-headedness has to be considered during the interpretation of a compound, because a change of the modifier and head can lead to a change of meaning, as illustrated by the example Orangensaft “orange juice”, and Saftorange “juice orange”, which are both lexicalized German words. German compounding is also characterized by its complexity. While most German compounds consist of two constituents, 10% include three elements, and 1.5% are made up of four constituents (Ortner and Müller-Bollhagen, 1991, cited in Schlücker, 2012). Compounds that contain particles like in überfließen “overflow” (Marchand, 1969 cited in Schlotthauer and Zifonun, 2008, p. 274) illustrate, that a distinction between syntactic and morphological units is somehow controversial.

In contrast to derivational morphology, which is a comparably clear rule-based operation, “compounding is governed by more malleable principles” (Leminen et al., 2019, p. 28). Compound word creation offers a certain degree of freedom in the placement of a particular constituent morpheme within a compound. Therefore, there are different possibilities for word constructions and two or more elements can be differently combined to create a novel compound (e.g., the constituent man can be the first or the last constituent, like in milkman or manpower, Leminen et al., 2019, p. 28). This freedom could represent a difficulty in acquisition.

Compounds are particularly fascinating from a linguistic perspective. On the one hand, they function as morphological units that can often be expressed through syntactic structures. On the other hand, they consist of two or more lexical elements that combine to form a single word, whose meaning may sometimes only be inferred from context. Additionally, phonological aspects are important when combining words, for example through changes of prosodical patterns (Grimm, 2010). Compounds therefore can be described as a linguistic interface phenomenon.

To date, studies on German-speaking children remain limited. Dressler et al. (2010) analyzed spontaneous speech data from two Austrian children and found that compounds first emerged at 1;8 for the boy and 1;10 for the girl. Initially, the boy produced lexical compounds such as Müllauto “garbage car” and Segelschiff “sailing boat” before creating novel compounds like Lasterwagen “truck car”. The girl, in contrast, began forming novel compounds such as Leniomi “Lenigranny” to refer to her grandmother. By 2;4, she was producing lexical compounds like Käsebrot “cheese bread” (Dressler et al., 2010, p. 326–327). The emergence of novel compounds is considered a sign of linguistic productivity. No instances of word order reversal were reported for the two children. However, the author describes such an error made by his daughter (2;6), who produced Berg-häferl instead of Häferl-berg for the name of a mountain, that looks like a cup (Dressler et al., 2010, p. 340).

In a longitudinal study, Schipke and Kauschke (2011) examined spontaneous speech from 39 mother-child interactions with monolingual German-speaking children aged 1;1 to 3;0. Their findings revealed both an absolute and relative increase in word formations within the second and third years of life, with compounding and derivation appearing simultaneously at around 1;9. Additionally, 10% of the observed word formations were lexical innovations, including compound nouns such as Automensch “car man” to describe a robot (Schipke and Kauschke, 2011, p. 75). Further observational research suggests that German-speaking children tend to create novel compounds to compensate for lexical gaps, particularly by using compound nouns. For instance, they might coin Brennlicht “burning light” to refer to stars (Stern and Stern, 1968, cited in Elsen and Schlipphak, 2015). The ability to form new words has been observed at a very early stage in German language acquisition

So far, further studies investigating the acquisition of compounds focused on semantic, phonological and morphological aspects. Lexical-semantic aspects of compounds are observed in studies focusing on the ability of understanding compounds (Krott and Nicoladis, 2005) and forming novel words with regard to the semantic relations that are conventional in their language (Clark et al., 1985; Nicoladis, 2003). Phonological abilities like prosodic word structure were investigated amongst others for Swedish (Mellenius, 1994) and German (Grimm, 2010). Morphological aspects such as the use of linking elements and inflection (Rosenberg and Mellenius, 2018) as well as plural use of the modifier (Nicoladis, 2005) were investigated for Germanic and Romance languages as well.

In this study, we will focus on the structural or morphosyntactic constraints of the formation of compound-nouns: the order of the head and the modifier. For the acquisition in English, there is evidence, that children understand the relation of the head and modifier at the age of two (Clark et al., 1985). This was investigated by applying a word-picture matching task with 60 children aged 2;0 to 6;0. In each trial, the child was presented with four pictures: one depicting the target word (e.g., mouse hat, a hat on a mouse), one showing only the modifier (mouse), one showing only the head (hat), and one acting as a distractor that included the correct head but a different modifier (fish hat, a hat on a fish). The authors interpret their findings as evidence that in English-speaking children a taxonomic category organization is already present in the mental lexicon by the age of two. However, it remains unclear whether children at 2;4 had already acquired the modifier head order in English, as it is uncertain how they would have interpreted reversed constructions such as hat mouse in this context (Clark et al., 1985). Clark et al. (1986) also investigated the production of novel compounds. Their findings suggest that 3- and 4-year-old children tend to produce more inverted, ungrammatical forms, such as puller wagon to describe someone who pulls a wagon, whereas 5- and 6-year-olds more frequently generate grammatical forms like wagon puller. This indicates a developmental phase in compound acquisition, during which English-speaking children initially invert the order of compound components. In a subsequent study, Clark and Barron (1988) asked children aged three to six to evaluate grammatical forms like wagon puller and ungrammatical forms like puller wagon and to correct them. Results showed a positive influence of age on the detection of ungrammatical forms. However, many of the children's corrections remained ungrammatical, which Clark and Barron (1988) interpreted as evidence that comprehension develops ahead of production.

However, there are indications from observational studies, that German-speaking children disobey the right-headedness and show inversions of the compound elements as well at a very young age around 2 years. (Elsen and Schlipphak (2015) report about observed inversions produced spontaneously by a 1;9 years old German-speaking girl referring to Fingernägel “fingernails” and Fußnägel “toenails” as Nagelfinger “nailfinger” and Nagelfuß “nailtoe”. Rainer (2010, cited in Elsen and Schlipphak, 2015, p. 2123) reports about a stage from 3 to 4 years, where children seem to have difficulties placing the constituents in the correct order. However, Dressler et al. (2010) found inversions to be rare in their study and Schipke and Kauschke (2011) do not mention any inversions at all. Nevertheless, Dressler et al. (2010) emphasize the relevance of investigating morpheme order in typologically different languages.

The studies mentioned above investigated exclusively monolingual children. Evidence for bilingual children on morphological acquisition is rare, but there are studies investigating cross-linguistic influence of English and French (Nicoladis, 2002), English and Persian (Foroodi-Nejad and Paradis, 2009) and Spanish and Japanese (Kutsuki, 2019) on production and reception of novel compound nouns with respect to the head modifier order. The studies revealed that bilingual children are more likely to invert compound components than their monolingual peers. This is interpreted as cross-linguistic influence, due to different word-formation strategies of the languages compared. The method of these three studies was applied to the German context by this research group (Scherger et al., 2024). They investigated 16 monolingual German-speaking children and 19 bilingual German-speaking children with 7 different first languages. The results support the previous findings that bilingual children might display cross-linguistic influence in applying the left-headedness when producing and understanding novel compounds. A higher inversion rate for production was also found for bilingual children compared to monolingual peers in another study by these author group (Scherger and Kliemke, 2021). They asked nine monolingual and nine bilingual children aged 7 to 8 years to name low frequent complex objects and analyzed their used word-formation strategies. The word-formation strategies observed did not differ in use and extent between bilingual and monolingual children.

Regarding the present investigation, the focus lies on DLD exclusively in monolingual acquisition in order to avoid a confounding of effects of DLD with effects that bilingualism might have. Because of this, bilingualism will be an exclusion criterion in this study, but investigated further in another study (Giesselbach and Scherger, submitted).

Developmental Language Disorders can affect semantic-lexical abilities, phonological abilities, morphological-syntactical and/or communicative abilities of children to an individual extent.

A delayed onset of vocabulary acquisition is often one of the first signs of DLD (Sansavini et al., 2021). As their development progresses, children with DLD typically fail to catch up with the vocabulary of their peers with typical development (TD). Research suggests that deficits in lexical-semantic organization may be a contributing factor (Sheng and McGregor, 2010). The constituents of compound words are often semantically related either through a part-whole relationship (meronymy) or a category-subtype connection (hyponymy), as illustrated by the example fruit tree. The hypernym tree functions as the head, while the modifier fruit provides further specification. Together, they form the compound noun fruit tree, which is a hyponym of tree. Deficits in lexical organization may therefore result in difficulties with both the comprehension and production of compounds. Additionally, compounding can result in complex word-formations, particularly in German, which, according to Motsch et al. (2022), may contribute to difficulties in word storage and retrieval. However, research has shown that children with lexical deficits create novel compounds to compensate for gaps in their vocabulary. In an intervention study by Ulrich (2012), cited in Motsch et al., 2022, p. 36), children referred to a Bügeleisen “iron” as Glattmacher “smooth maker”. This demonstrates the formation of a synthetic compound that describes the function of the object as a reaction of a lexical gap. Grimm (2010) investigated the interaction between phonological and lexical abilities in German compounding as part of her dissertation. She describes that an increase in lexical knowledge drives the phonetic differentiation of words, which in turn enhances the phonological precision of target word production. Words are stored in connection with phonologically similar words, known as phonological neighbors (a. o. Yates, 2004). To differentiate phonological neighbors from one another, additional storage of phonetic details is required, which supports the computation of phonotactic probabilities. In German, certain phoneme sequences, such as /nt/ in Hand (“hand”), are more common, so called phonotactically legal, whereas sequences like /sk/ in Skat aren't common or phonotactically illegal. Compounding often leads to phontactically illegal phoneme sequences, such as /tb/ in Luftballon (“air balloon”; Grimm, 2010). Because of this, Grimm argues that the constituents of compounds are stored separately, which could lead to inversions of word order. How these acquisition processes manifest in children with DLD remains unclear and is therefore identified by Grimm (2012) as a research desideratum and has not been researched since.

As compounding is defined as a linguistic interface phenomenon, compounding could be affected by children with DLD on different linguistic levels.

To date, relatively few studies have examined compounding in children with DLD. Most of the studies are from English speaking surroundings, mostly focusing on morphological aspects, like inflection and plural formation in synthetic compounds (van der Lely and Christian, 2000; Clahsen and Almazan, 2001), indicating that children with DLD have more difficulties than their peers with TD. Three studies focused on phonological aspects of compounding, reporting difficulties for children with DLD who speak Greek (Dalalakis, 1999; Kehayia, 1997) and Japanese (Fukuda and Fukuda, 1999). Padrik (2005) investigated 40 Estonian children with and without DLD. They were asked to name people who carried out different activities, like the man catches fish. In Estonian, this can be realized as a conversion fisher, a compound-noun fishman or a synthetic compound fish-catcher. Results showed that children with DLD were more likely to produce phrases and simple words like man to describe the people, compared to the children with TD, who produced mostly conversions, which is according to Padrik (2005) the preferred word-formation in Estonian for personal descriptions. A disorder that frequently co-occurs with DLD is autism spectrum disorder (ASD). The comprehension of conventional compounds was examined by Riches et al. (2012) in four groups: adolescents with TD, adolescents with ASD with and without Language Impairment (LI), and adolescents with DLD. Adolescents with ASD-LI and DLD exhibit similar difficulties in explaining the meaning of compounds, which has been attributed to semantic-lexical deficits. Additionally, Kambanaros et al. (2019) found that children with ASD had more difficulty explaining the meaning of novel compounds compared to their TD peers. In line with the research focus of the present study, Grela et al. (2005) and McGregor et al. (2010) found more inversions in children with DLD than in children with TD. Grela et al. (2005) asked ten English-speaking children aged 4;8 to 7;0 to invent new names for example for animals living in a certain place or objects that are made of a certain material. All children formed compounds, but only 65% of the children with DLD formed right-headed compounds. Children with TD followed the principle of right-headedness in 97% of the utterances, which indicates a difficulty for children with DLD compared to children with TD applying the right-headedness when producing novel compounds. Following the methodical approach by Grela et al. (2005), McGregor et al. (2010) found comparable results, as children with DLD (aged 5;0 to 8;6) showed more word error mistakes when producing novel root compounds (e.g., umbrella horse for an umbrella with horses on it). However, children of both groups produced novel compound-nouns. Additionally, they asked children with DLD to explain conventional English compound-nouns (e.g., “Why do we say car door?”), like (Krott and Nicoladis 2005) did. The children with TD outperformed children with DLD at explaining the meaningful relationship between the modifier and head. The authors attributed this difficulty to faulty lexical-semantic processing abilities in children with DLD related to reduced knowledge of the relationships between words in their semantic lexicons.

According to Libben et al. (2020), compounds provide valuable insights into the organization of the mental lexicon due to their dual nature as both integrated structures and combinations of recognizable sub-elements. Thereby the question arises, whether compounds are processed holistically or decomposed. There is evidence for multiple routes from EEG, MEG- and fMRI research (MacGregor and Shtyrov, 2013; Holle et al., 2010; Leminen et al., 2019). The psychocentric view suggests that both whole-word representation and representations of constituent lexemes are necessary to process compounds (Libben, 2014). The argument of efficiency can be used in favor for both processing ways: representing only single constituents of compounds in the mental lexicon would result in considerable storage efficiency, while representing all multimorphemic compounds in their full forms would increase the size of the mental lexicon but result in less computational parsing mechanism needed (computational efficiency, see Libben, 1998). Leminen et al. (2019, p. 31) point out, that “research on compounding is somewhat scarce” and that further research is needed. (Semenza and Mondini 2015) emphasize, that word-formation investigations in language disorders reveal interesting information about processing of complex words. To the best of our knowledge, no studies have investigated language processing in relation to compounding in children with DLD.

To summarize, German-speaking children begin using compounds before the age of two, primarily relying on lexicalized words but also creating novel compounds already (Dressler et al., 2010; Schipke and Kauschke, 2011). Early in language acquisition, difficulties in applying right-headedness can lead to the production of inverted compounds (Elsen and Schlipphak, 2015). Research on English-speaking children with DLD suggests that they struggle more with right-headedness when producing and understanding conventional and novel compounds compared to their peers with TD (Grela et al., 2005; McGregor et al., 2010). However, there is limited evidence on the compounding competencies of German-speaking children with DLD and the implications for their processing of complex words. This study addresses this gap by focusing on German-speaking children aged 4 to 5 years. At this age, children with TD have typically mastered the principle of right-headedness, whereas English-and Estonian-speaking children with DLD show persistent difficulties (Padrik, 2005; Grela et al., 2005; McGregor et al., 2010).

To compare the word-formation strategies in TD and DLD, 61 monolingual children with DLD (n = 30) and TD (n = 31) aged 4 to 5 years were recruited from speech therapy practices and kindergartens in North-Rhine Westphalia, Germany. The presence of typical language development was verified using the Sprachentwicklungstest für 3-5-jährige Kinder (SETK 3-5, Grimm, 2015). None of the TD children were receiving speech or language therapy at the time of the study. The results of the SETK 3-5 confirmed no need for therapy. Furthermore, children with other disorders, likely to co-cause a DLD, were excluded from the study. Therefore, parents confirmed that their children were not suspected or diagnosed with the following conditions, which are according to Bishop et al. (2017) and Lüke et al. (2023) considered to be other contributory causes of DLD: intellectual disabilities (IQ below 70), genetic syndromes, hearing disorders, childhood aphasia, childhood brain damage, neurodegenerative diseases, disorders from the autistic spectrum, and motor disorders. All children with DLD were already receiving speech and language therapy on average for 13.7 months (SD = 6.0, range = 3–28 months). The therapy duration for eight of these children is unknown. Information regarding therapy duration, therapeutic focus and diagnostic assessments was obtained from the treating speech-language therapists with a written questionnaire. According to the overarching nature of the phenomenon compounding as outlined above, the children's therapy focus had to include at least one of the following structural linguistic domains: lexical-semantics, morphology-syntax, or phonology. Children with isolated phonetic disorders, fluency disorders, or voice disorders were excluded from the study. For 12 children, therapy focus was on lexicon-semantics and for 16 children it was on phonology, while only two children received therapy due to deficits in the morphosyntactic domain. According to the therapists, the diagnoses were made using the following standardized, normed language tests from German: SETK3-5 (Grimm, 2015), PDSS (Kauschke et al., 2022), TROG-D (Fox-Boyer, 2023) or AWST-R (Kiese-Himmel, 2005). For 19 children, no information is available on the assessment tests used for the diagnosis. A total of 31 children with TD was assessed twice in their kindergarten setting: first, for the exclusion of DLD, and on a separate session for the experimental tasks. Testing took place in a quiet room, where the experimenter presented visual stimuli on a laptop. The same experimental tasks were applied to 30 children with DLD within their speech therapy practices. However, due to advantages in terms of practicability, 13 assessments were conducted remotely via Zoom, with the experimenter connecting to the speech therapy practice online and receiving on-site support from the speech therapist. As Table 1 shows, the age ranges are comparable between groups (Mann-Whitney U = 300.00, p = 0.340).

Additionally, we analyzed adult data from two pilot studies (Scherger and Kliemke, 2021; Scherger et al., 2024), including 33 adults aged 18 to 73 years. The adults were asked to name the low-frequency objects (experiment 1) or novel compounds (experiment 2) in a paper-pencil study. Table 1 provides an overview of the participant characteristics.

To explore the word formation strategies of the two groups, two different experimental procedures were applied.

The aim of the first research question was to investigate the word-formation strategies when naming low frequency objects of children with DLD compared to peers with TD and adults. In order to answer the first research question, we compared the percentages of the analyzed responses, with a particular focus on compounds, including their inversions and complexity. Due to the unequal distribution of data across the three groups, we applied the non-parametric Kruskal-Wallis test.

Addressing the second research question, the inversion rates were compared in the production task between the three groups. Due to non-equally distributed data, we calculated the rank sum Kruskal-Wallis test as well. For the reception task, we compared the inversion rates between the two child groups. Since the data were not normally distributed, we applied the Mann-Whitney U test.

With regard to the third research question, the influence of therapy focus and duration on the performance of the children with DLD in the production and reception task of experiment 2 was analyzed calculating two multiple linear regression models.

In order to answer the fourth research question, multiple linear regressions were carried out to investigate the influence of age, gender and DLD on the inversion rate in the production and reception task of experiment 2.

To account for multiple comparisons, we applied the Bonferroni correction to control the family-wise error rate. Given that four hypotheses were tested simultaneously, the conventional significance level of α = 0.05 was divided by the number of tests (4), resulting in a corrected alpha level of α = 0.0125 for each individual test.

The calculations were carried out with SPSS (IBM).

In total, 786 responses were analyzed across the three groups. Three children with DLD and one child with TD were excluded from the analysis as they did not respond to at least half of the items. Among the remaining participants (TD: n = 27, DLD: n = 25, adults: n = 10), 10.9% of responses were zero reactions. Figure 4 provides an overview of the analyzed responses from the children with TD and DLD and the adult control group. Across all groups, compounds were the most frequently used strategy for naming the low-frequency objects (TD: M = 55.96, SD = 24.84; DLD: M = 59.52, SD = 21.72; adults: M = 69.63, SD = 15.05). On a group level, 43.0% of the produced compounds were synthetic compounds. The second most common response type was classified as simplizia (TD: M = 22.45, SD = 18.53; DLD: M = 20.14, SD = 14.04; adults: M = 18.54, SD = 12.92) while derivation was the least frequent word-formation strategy (TD: M = 4.59, SD = 8.02; DLD: M = 5.84, SD = 9.04; adults: M = 0.76, SD = 2.30). On average, 13.4% of the items were described in a phrase.

After Bonferroni correction, there was no significant difference for zero reactions between the three groups (Kruskal-Wallis-H = 6.5, p = 0.038). For the other reaction types, no significant differences were observed across the three groups (Kruskal-Wallis-test: compounds: H = 2.7, p = 0.175; simplizia: H = 0.0, p = 0.957; derivations: H = 2.9, p = 0.223; phrases: H = 1.1, p = 0.564).

In their compound production, five children with DLD, four children with TD, and one adult inverted compound components. The overall inversion rates were 3.8% for children with TD, 3.2% for children with DLD, and 1.0% for the adult control group. A Kruskal-Wallis test showed no significant difference between the three groups (H = 0.4, p = 0.780). Across all groups, the majority of compounds consisted of two components (TD: 72.1%, DLD: 77.8%, adults: 55.3%) like Blasenstab “bubble wand”. Compounds consisting of three elements, such as Seifenblasenstiel “soap bubble stick”, were produced in 25.9% of cases by children with TD and in 20.4% of cases by children with DLD, whereas adults formed 32.7% of their compounds with three components. More complex compositions with four or more components were rare in both child groups (TD: 2.0%, DLD: 1.8%) but occurred descriptively more frequently in adults (11.4%). However, children in both groups also formed complex compounds. Examples include Drunterhertauchband “underneath diving band”, produced by a child with TD for Item 6, and Kugelschreibereinfahrer “ball point pen retractor” produced by a child with DLD for Item 5. A Kruskal-Wallis test confirmed no significant differences between the three groups in terms of compound complexity (two components: H = 3.4, p = 0.174, three components: H = 2.0, p = 0.366, four components: H = 2.4, p = 0.299).

The two multiple linear regression models assessing the effects of therapy focus and duration on inversion rates in the production and reception tasks of experiment 2 for children with DLD revealed no significant predictors, as shown in Table 3.

Two multiple linear regression models were calculated to investigate the influence of age, gender and DLD on inversion rates in the production and reception task of experiment 2 for all children.

As you can see in Table 4, there were no significant influences of gender and age on inversion rates in the production and reception task for both groups. However, DLD is a statistically significant predictor in the reception task.