The issue of writing words in languages as either adjacent or separate by means of orthographic symbols, in other words, writing with spaces in between, shows variations both among different languages and within the same language. It is noted that at the beginning, period (.) was used to separate words, and after the 7th century A.D., the habit of separating words using spaces became evident (Başkan, 2003). Even though compound words exhibit universal common features morphologically, syntactically, and semantically, the orthographic spaces between the components of a CW may not display a distinct pattern. Notably, there is an ambiguity in the spelling of compound words both within and across languages: some compound words are written together, while others are written separately. For instance, in German, a compound word like "Kinderbuchclub" does not require orthographic spaces, whereas in Turkish, a writing form like *çocukkitapkulübü is considered incorrect. Although word boundaries and how they are reflected in spelling are issues concerning a language, matters related to the spelling of compound words are often excluded from linguistic debates. This study will discuss the fundamental phonological, morphological, semantic, and syntactic characteristics of compound words in Turkish instead of providing a detailed classification. It will examine whether there is a pattern in the tendency of language users while writing compound words with a space or adjacent, based on these CWs’ linguistic features. 

Research Question:

Is there a pattern in the writing of compound words in Turkish based on a) semantic transparency/opacity, b) the lexical category of components, and c) the parameter of centrality?

Research Sub-Questions:

1. Does the transparency or opacity of the meaning to which the compound word refers (see 3a-c) significantly affect users’ orthographic tendencies?

2. Does the lexical category of the elements constituting the compound word (see 4a-f) significantly affect users’ orthographic tendencies?

3. Does the presence of the semantic head inside or outside the compound word—the centrality issue - (see 6a-c) significantly affect users’ orthographic tendency?

4. Does the formation of a compound word with the compound marker–(s)I significantly affect the users’ orthographic tendency?

In this study, a quasi-experimental design, one of the quantitative research methods, was used. An offline lexical decision task (LDT) was administered to participants via Google Forms for data collection. The test items used in the task are listed in Appendix 1. All the test items are newly coined compound words that do not exist in Turkish but can be compounds. Participants were first presented with an online Consent Form and asked to fill in their education, age, and gender information. Afterwards, in Test 1, the participants were asked to decide whether the given compound word definition should be written adjacent or with a space (see 9a). Those who completed Test 1 were then directed to Test 2. In Test 2, the compound words defined in Test 1 were presented within sentences, and participants were asked to mark "Correct" if they found the spelling (in terms of adjacency or separation) correct, or "incorrect" if they found it wrong (see 9b).

The test items presented to the participants were labelled based on semantic transparency/ opacity, semantic headedness, and the lexical categories of the components of CW. Subsequently, a 3 (endocentric, exocentric, coordinate) X 4 (noun+noun, noun+noun-(s)I, noun+verb, verb+verb) X 2 (semantically opaque, semantically transparent) mixed-design ANOVA (repeated measures mixed ANOVA) was used to analyse the data from the participants’ responses in the study. For post hoc multiple comparisons, the Bonferroni correction was applied.

According to the descriptive statistical results (see Table 2), it has been found that semantically opaque words (e.g., karga.tüy, kamçı.bacak; see Appendix 1A) tend to be written together more frequently compared to semantically transparent compounds (66.4% and 50.6%, respectively). To determine the statistical significance of this percentage difference, a repeated measures ANOVA was applied. It was found that the difference between semantically opaque and semantically transparent compound words was statistically significant (F=7,975, p<.001). In other words, semantically opaque compound words significantly increase the tendency to be written together. Additionally, compound words formed with the compound marker –(s)I significantly influenced the writing tendency (whether to be written adjacent or with a space) of semantically opaque and transparent compounds (F=7.786, p<.001). According to the pairwise comparison results (see Table 3), the percentage difference in the writing of transparent-∅ and opaque-∅ compound words (72.8% and 78.4%, respectively) was not statistically significant (p=.788), indicating that participants tended to write both semantically opaque and transparent words together when there was no compound marker –(s)I. However, the percentage difference in the writing of the transparent and opaque compounds formed with the compound marker was statistically significant (28.4% and 54.4%, respectively) (p<.001). In summary, semantically transparent compound words formed with the CM -(s)I significantly increase the tendency of participants to write them separately.

When examining how the lexical category of the components of a CM (noun+noun, noun+noun-(s)I, noun+verb, and verb+verb) affects the writing tendency of native Turkish speakers to write them together or separately, it is observed that compounds formed with noun+noun components have the highest tendency to be written adjacent at 75,6%. The formations with verb + verb components increase the tendency to be written together at a rate of 74,5%, and those from noun + verb components at a rate of 69,6%. Compound words of the noun+noun-(s)I type, formed with the compound suffix, result in a higher tendency to be written separately with a rate of 41.4% (see Table 2). The percentage difference between compound words belonging to different word meanings was statistically significant and the effect size was high (F=4.399, p<.001). Pairwise comparisons were performed to determine the reason for this significant difference. According to the results of the pairwise comparisons (see Table 3), there was no statistically significant difference between the compound words with the N+N and V+V components (p=1.000), and no significant difference between the compound words with the N+V and V+V components (p=.187). However, it was observed that the percentage difference between the compound words formed with the CM , N+N-(s)I and other compound word groups (N+N, N+V, V+V) was statistically significant (p<.05). Therefore, it can be stated that the components of the compound words affect the tendency of Turkish native speakers to write them adjacent or separately, and the presence of the CM –(s)I structure increases the tendency to put an orthographic space between the words among participants.

Finally, the study examined how the presence of semantic headless within (endocentric) or outside (exocentric) the compound word or the presence of a compound word being bidirectional (coordinate/two-headed) influenced the tendency of Turkish native speakers to write the words adjacent or separately. Descriptive statistics show that external-headed compounds (e.g., arpa. göz (disease), timsah.beli (fruit), teker.yürür (vehicle); see Appendix 1A) resulted in the highest tendency to write together at a rate of 74%. Participants showed a tendency to write together with coordinate (two-headed) compound words (e.g. çırp.uç, yakar.akar; see Appendix 1A) at a rate of 66.4%, but this tendency decreased to 50.6% in endocentric words (e.g. kedi. bilimi, see Appendix 1A). The percentage difference between the endocentric, exocentric, and coordinate compounds was found to be statistically significant and had a high effect size (F=6.194, p<.001). According to the results of pairwise comparisons to determine the reason for this difference (see Table 3), the difference between the endocentric compound words and both the exocentric and coordinate compounds was statistically significant (p<.001). 

The findings can be summarised as follows:

1. Semantically opaque compound words increase the tendency to be written adjacently, while semantically transparent ones increase the tendency to be written with an orthographic space.

2. Forming compound words with the CM -(s)I significantly increases the tendency to be written separately in both opaque and transparent compounds.

3. Exocentric compounds increase the tendency to be written separately.

4. Having at least one or both components of the compound word belonging to the verb category significantly increases the tendency of adjacency.

5. Components belonging to the noun+noun category, formed with the CM -(s)I, semantically transparent and endocentric compound words are the test elements that cause the highest tendency to write separately, while opaque and exocentric compound words are the test elements that cause the highest tendency to write together.

Based on these data, it can be seen that in the writing process of compound words, both the concept referred to by the compound word and the relational meaning between the components/ elements of the compound are constantly active. It can be said that the compound marker –(s) I may strengthen the semantic relationship between the components regardless of whether the concept referred to by the compound is opaque or transparent, and therefore leads to the tendency to write them adjacently. However, the tendency of the coordinate (two-headed) compounds to be written together may be due to the violation of the semantic hierarchy, since in Turkish, the element that contributes more to the meaning (semantic head) is located on the left side. Lastly, the tendency of having at least one component, which is the verb stem, increases the tendency to write adjacently (e.g. N+V: teker.yürür; V+V: çırp.uç) can be seen as a tendency to reduce ambiguity, emphasising that the speaker's perspective is at the word level, not at the sentence level, which is the structure of this compound word. Differences in the orthography of compound words across languages (leaving spaces during the writing process) might provide clues as to how syntactically related word groups (chunks) are memorised differently by speakers of different languages and answer questions such as "What is a word?" or "What are the boundaries of words?". Because compound words in Turkish are productive, new compounds will continue to be produced. Both intra-lingual differences among the same language speakers and inter-lingual differences should be considered regarding whether compound words are written together or separately. Therefore, instead of determining inflexible criteria for spellings, a descriptive approach that accepts the ambiguous nature of these words while being written, where spelling is influenced by formal, semantic, and syntactic variables simultaneously, could be followed.