JĘZYKOZNAWSTWO nr 3(28)/2026
ISSN 2391-5137
https://doi.org/10.25312/j.10698
Anna Yang https://orcid.org/0009-0007-5825-8871 School of Chinese Language and Literature
Beijing Foreign Studies University e-mail: anna.kun.yang@gmail.com
Why ‘what’ is harder than ‘where’
in Polish-Mandarin bilingualism: Complexity and selective cross-linguistic influence
in the development of wh-questions1
This article examines the acquisition of wh-questions in a Polish-Mandarin bilingual child (2;0-3;1) with a focus on how computational difficulty shapes cross-linguistic influence. The analysis draws on a longitudinal corpus of naturalistic interactions and targets the contrast between ‘what’ (Polish co, Mandarin shénme) and ‘where’
1 This study was made possible by the financial support provided by the Confucius China Studies Program, a fellowship administered by the Centre for Language Education and Cooperation. I extend my deepest gratitude to my PhD supervisor, Professor Manchun Dai, for unwavering guidance and feedback throughout this research. Special thanks to my family for their invaluable support – particularly my daughter, whose participation made this longitudinal study possible.
(Polish gdzie, Mandarin nǎlǐ) questions in Polish, Mandarin, and Polish-matrix (code-mixed) utterances. Two complexity measures are applied: (1) a movement-based derivational index and (2) a feature-reassembly index that incorporates the feature inventory, constructional markedness, and meaning-form mapping burden. The results reveal a systematic wh-type asymmetry that is not captured by dominance alone. In Polish-directed production, ‘what’ questions show stronger Mandarin influence, including frequent use of Mandarin wh- lexemes and in-situ configurations, whereas ‘where’ questions more often display Polish-like fronting even when Mandarin lexical material is used. In Mandarin-directed production, residual Polish influence is more evident in the ‘where’ domain than in the ‘what’ domain. These patterns align with the feature-reassembly metric: the Polish ‘what’ domain imposes a higher mapping burden due to case-conditioned alternants, while the ‘where’ domain remains morphologically stable. The findings support a complexity-sensitive account of selective cross-linguistic influence in bilingual wh-development.
Keywords: bilingual first language acquisition, wh-questions, computational complexity, cross-linguistic influence, Polish, Mandarin Chinese
Artykuł dotyczy przyswajania pytań szczegółowych przez dziecko dwujęzyczne polsko-mandaryńskie (2;0–3;1) i pokazuje, jak złożoność kształtuje wpływ międzyjęzykowy. Analiza opiera się na longitudinalnym korpusie naturalnych interakcji. Badanie koncentruje się na kontraście między pytaniami typu „co” (pol. co, mand. shénme) i “gdzie” (pol. gdzie, mand. nǎlǐ) w wypowiedziach polskich, mandaryńskich oraz w wypowiedziach z polską ramą zdaniową i wtrąceniami z mandaryńskiego (code-mixing). Zastosowano dwie miary: (1) zło- żoność derywacyjną (liczbę zależności przestawieniowych) oraz (2) złożoność reorganizacji cech (inwentarz cech, markowanie konstrukcyjne i obciążenie relacji znaczenie–forma).
Wyniki pokazują wyraźną asymetrię, której nie wyjaśnia sama dominacja językowa. W wypowiedziach ukie- runkowanych na polszczyznę pytania typu „co” częściej przejmują rozwiązania mandaryńskie (wyrazy pytajne z mandaryńskiego i realizacje in situ), natomiast pytania typu „gdzie” częściej zachowują polską strategię frontingu nawet przy użyciu mandaryńskiego materiału leksykalnego. W wypowiedziach ukierunkowanych na mandaryński wpływ polski jest bardziej widoczny w domenie „gdzie” niż „co”. Wzorzec ten jest zgodny z ujęciem reorganizacji cech: polska domena „co” wiąże się z większym obciążeniem mapowania (m.in. przez alternacje przypadkowe), podczas gdy domena „gdzie” jest morfologicznie stabilna.
Słowa kluczowe: dwujęzyczne przyswajanie pierwszego języka, pytania wh, złożoność obliczeniowa, inter- ferencja międzyjęzykowa, język polski, język chiński (mandaryński)
Bilingual development is rarely uniform across the grammar. Even within a single child, some constructions stabilize early and remain robust, whereas others show prolonged variability, non-target options, or systematic cross-linguistic influence (Hulk, Müller, 2000; Sorace, 2011, 2012; Meir, Janssen, 2021). This selectivity raises a general explanatory question for bilingualism research: what determines where bilingual interaction concen- trates, beyond the fact that exposure and use are often unequal across the two languages? One widely used predictor is language dominance, often operationalized via input and output balance. Dominance shifts can coincide with rapid gains, temporary regressions, and changes in mixing frequency, yet dominance is frequently underdetermining for mor- phosyntactic outcomes (Paradis, 2011; Armon-Lotem, Meir, 2018). This motivates treat- ing bilingual development as shaped by an interplay between language-external factors (including dominance) and language-internal constraints that render some structures more
difficult to compute, retrieve, or map to overt forms (Meir, Janssen, 2021).
Even in domains where typology and dominance would seem to favor transfer, the em- pirical record is mixed: some bilingual and L2/heritage studies report cross-linguistic transfer in wh-questions (hereafter, wh-transfer) toward the less costly option, while others find little or no transfer under comparable conditions. This has motivated accounts in which dominance and typology create opportunities for cross-linguistic influence, but structural and ecological constraints determine whether transfer is realized and where it concentrates (Strik, 2012; Park-Johnson, 2017; Qi, Di Biase, 2020).
One influential language-internal constraint is computational difficulty. Complexi- ty-based approaches predict selective cross-linguistic influence: when competing gram- matical options are available, learners tend to rely on options with lower derivational cost and/or lower feature-mapping burden, especially under competition between languages (Lardiere, 2008, 2009; Slabakova, 2009; Jakubowicz, 2011; Meir, Walters, Armon-Lo- tem, 2017). Feature reassembly work on child heritage grammars likewise suggests that difficulties in mapping grammatical features onto their overt morphological forms can lead to persistent optionality and mismatches between morphology and syntax (Cuza, Pérez-Tattam, 2016).
This article explores these predictions in a typologically distant pairing (Polish-Man- darin) using a longitudinal corpus of one bilingual child (2;0–3;1). Interrogatives provide a useful diagnostic domain because they combine typological contrasts (e.g., wh-place- ment) with morphosyntactic dependencies that vary sharply across languages and are known to be sensitive to complexity (Jakubowicz, Strik, 2008; Strik, 2009; Jakubowicz, 2011). Earlier work on the same dataset emphasized dominance shifts in the emergence and consolidation of interrogative patterns (Yang, 2025). Here I adopt the same de- velopmental backdrop but reframe the explanatory target: rather than asking whether dominance matters, I ask why two closely related subdomains respond differently under similar dominance ecology.
The analysis focuses on the contrast between ‘what’ and ‘where’ questions: Polish co vs. Mandarin shénme, and Polish gdzie vs. Mandarin nǎlǐ. These two wh-domains are su- perficially similar (both are wh-questions), but they differ in the feature and morphological burdens imposed by Polish, particularly once case-related realizations and lexicalization choices are considered (Lardiere, 2008, 2009; Meir, Walters, Armon-Lotem, 2017). This contrast therefore functions as a minimal test of whether dominance-only predictors suffice or whether a complexity-sensitive account is required.
The paper addresses three questions: (RQ1) How do ‘what’ and ‘where’ questions differ in wh-position, code-mixing patterns, and the persistence of cross-linguistic influ- ence across three input phases? (RQ2) To what extent can this asymmetry be predicted by dominance measures alone versus by derivational and feature-reassembly complexity profiles of the relevant constructions? (RQ3) Do complexity-sensitive effects manifest similarly at onset and at later maintenance after dominance shifts?
In bilingual acquisition research, language dominance (often operationalized via input quantity/quality, use patterns, or proficiency measures) is frequently invoked to predict the direction of cross-linguistic influence (CLI) (Deuchar, Muntz, 2003; Yip, Matthews, 2006; Kupisch, 2008). Yet, dominance does not always entail global transfer: bilingual grammars may develop largely autonomously in some domains while showing transfer or optionality in others (Paradis, Genesee, 1996; Hulk, Müller, 2000; Müller, Hulk, 2001). This selectivity has motivated accounts that emphasise structural overlap, interface vul- nerability, and processing limitations (Sorace, Serratrice, 2009) rather than dominance alone. More generally, research on bilingual first language acquisition highlights early differentiation alongside the possibility of mixing and co-activation at the level of use (Lanza, 1997; De Houwer, 2009).
The present analysis treats dominance as a necessary but insufficient condition. Dom- inance can create conditions for dormancy (temporary non-use) or activation (renewed productive use) of constructions, especially when input conditions change (Montrul, 2016). However, the wh-type asymmetry discussed below arises within the same time windows and thus calls for a factor that distinguishes ‘what’ and ‘where’ beyond the broad dominance profile.
Computational complexity has been used as a predictor of acquisition difficulty in ap- proaches that link grammatical operations to processing demands. In the Derivational Complexity Hypothesis (DCH), structures that require more syntactic operations – espe- cially additional movement dependencies – are expected to be acquired later and to be more vulnerable under demanding conditions (Jakubowicz, Strik, 2008; Jakubowicz, 2011). Work on bilingual wh-development has used this logic to motivate selective cross-linguis- tic influence: when two systems provide competing derivational options, learners may preferentially rely on the derivationally less costly option (Strik, 2012).
A complementary line of work locates difficulty not only in derivation, but in the map- ping between features and their overt realizations. In Feature Reassembly approaches, acquisition involves assembling and stabilizing feature bundles and their exponents across non-isomorphic systems (Lardiere, 2008, 2009). Chinese scholarship has proposed explicit ways of quantifying this mapping burden (Dai, 2011; Liu, Dai, Li, 2013; Mao, Dai, 2015; Mao, 2020), and related work links increased computational demands to limits on transfer in interrogative constructions (Yuan, 2017). Such representational challenges have also been invoked in models of heritage and bilingual grammars where reduced input or competing systems affect the stability and accessibility of feature configurations (Benmamoun, Montrul, Polinsky, 2013; Putnam, Sánchez, 2013).
These two perspectives converge on a shared prediction: cross-linguistic influence should be selective, and selectivity may emerge not only between languages, but also
within narrowly defined domains when competing options differ in derivational cost and/ or mapping burden.
Mandarin Chinese is typically characterized as a wh-in-situ language: wh-expressions such as shénme ‘what’ and nǎlǐ ‘where’ generally remain in their base positions, with interrogative scope interpreted at the interface (Huang, 1982; Cheng, 1997). Polish, by contrast, is a wh-fronting language: content questions canonically place the wh-phrase clause-initially, and wh-fronting interacts with other clause-structural dependencies, in- cluding verb-placement patterns and the positioning of the reflexive clitic się.
From a complexity perspective, this contrast suggests a clear derivational asymmetry between the two systems, with Polish requiring additional movement operations. It mo- tivates expecting CLI in bilingual development, but it does not predict within-domain selectivity. To probe such potential variation, the present study focuses on two frequent wh-items: co ‘what’ and gdzie ‘where’. These are derivationally comparable, as both canonically undergo wh-fronting and may involve similar movement configurations. The crucial difference lies in their morphosyntactic profiles. The co-domain is associated with a case-conditioned paradigm (e.g. co, czego, czemu, czym), whereas gdzie is largely morphologically invariant, yielding a lower mapping burden.
This contrast provides a minimal testing ground for complexity-based predictions: if computational complexity plays a role, differences in feature load and mapping should result in asymmetric susceptibility to cross-linguistic influence, even within the same construction type.
The data come from a longitudinal naturalistic study of one simultaneous Polish-Manda- rin bilingual child (pseudonym: M). The child was recorded regularly in home settings between 2;00 and 3;01. The corpus consists of 38 audio/video recording sessions; ap- proximately 20-minute segments with relatively dense and continuous child speech were selected from each for transcription and analysis. The selected segments were transcribed and coded for language (Polish, Mandarin, mixed) and interrogative type.
The participant is the author’s daughter. Written informed consent for the recording, transcription and academic use of the anonymized data was obtained from the other par- ent and other participating family members; identifying information was removed, and recording was discontinued whenever the child was unwilling to participate. The dual parent-researcher role may introduce interpretive bias, although the use of recorded and systematically coded data rather than retrospective observation was intended to reduce this risk.
Across the corpus, 475 interrogative utterances were identified, including 244 wh-ques- tions. Within the wh-domain, the most frequent items relevant to the present analysis are shénme ‘what’ (64 tokens) and co ‘what’ (56 tokens), as well as nǎ ‘where/which’ forms (40 tokens) and gdzie ‘where’ (22 tokens). These distributions make it possible to compare ‘what’ and ‘where’ trajectories within the same child and sampling frame.
Following the dominance-oriented analysis in Yang (2025), the child’s development can be divided into three broad phases based on use patterns and measures of morphosyntactic growth. Stage 1 (approximately 2;00–2;04) is Chinese-dominant; Stage 2 (approximately 2;05–2;09) is Polish-dominant; Stage 3 (approximately 2;10–3;01) shows a return to Chi- nese dominance. The present paper treats these stages as the macro-context in which wh-patterns unfold, while probing wh-type asymmetries that cut across the same stages.
Wh-questions were coded for (i) language context (Polish-only, Mandarin-only, code- mixed), (ii) wh-type (‘what’ vs ‘where’), and (iii) wh-placement (fronted vs in situ). The main focus is Polish-matrix code-mixed questions, defined as Polish clausal frames containing at least one Mandarin element (typically a wh-lexeme/phrase). These utterances are analysed as single productions reflecting competition between Polish wh-fronting and Mandarin wh-in-situ options (Goldrick, Putnam, Schwarz, 2016), allowing a direct test of whether the child maintains the frame-language requirement or follows the embedded wh-lexeme’s placement pattern (Yip, Matthews, 2000; Yip, Matthews, 2007).
This study operationalizes two measures for the present dataset. To model within-domain selectivity, the analysis introduces feature reassembly complexity (RCF), defined as RCF = NF + CF + MF. NF is the number of feature types assumed to be active for wh-in- terrogatives in the relevant system; CF captures construction-level marked requirements; MF captures the mapping burden between morphosyntactic requirements and overt real- ization (Dai, 2011; Liu, Dai, Li, 2013; Mao, Dai, 2015; Mao, 2020).
Under the feature inventory adopted here, Polish wh-questions involve [+wh], [+Q], [+Focus], [+EPP], [+Case] (NF = 5), whereas Mandarin involves [+wh], [+Q], [+Fo- cus], [+Topic] (NF = 4) (Huang, 1982; Cheng, 1997; Chomsky, 2001; Gębka-Wolak, Walkiewicz, 2013; Hu, 2019). MF is operationalized as the number of distinct overt forms among which a wh-lexeme must select in target production: invariant forms are set to MF = 1 (e.g., gdzie; shénme/nǎlǐ), while case-conditioned alternations increase MF (here: co/czego/czemu/czym → MF = 4). CF encodes configuration-level markedness as an additive measure: CF = 1 for in-situ wh and CF = 2 for canonical wh-fronting. Within Polish wh-fronting, the co-domain receives one additional point because fronting is com- bined with case-conditioned realization, yielding CF = 3, whereas invariant gdzie remains at CF = 2. In tokens with additional coded clause-structural dependencies (V2-like verb displacement and/or non-default się), CF is incremented by +1 in total (tab. 1).
The resulting values are used as an ordinal profile rather than as psycholinguistic magnitudes; they generate the prediction tested in section “Results” that Mandarin-like in-situ placement should be disproportionately attractive in the higher-mapping-load Pol- ish ‘what’ domain (co, RCF=12~13) relative to the lower-mapping-load Polish ‘where’ domain (gdzie, RCF=8~9).
Tab. 1. Complexity profile of the target wh-constructions (DC and RCF)
Construc- tion | Canonical wh strategy | DC | NF | CF | MF | RCF (approx.) | Notes |
Mandarin shénme / nǎlǐ | in situ | 1 | 4 | 1 | 1 | 6 | invariant wh; in-situ |
Polish gdzie | fronting | 2 (+1/2) | 5 | 2 (+1) | 1 | 8~9 | invariant wh; fronting |
Polish co | fronting | 2 (+1/2) | 5 | 3 (+1) | 4 | 12~13 | case paradigm increases MF (number of overt alter- nants) |
Note: In coded Polish configurations, V2-like verb displacement and/or a coded non-default się dependen- cy contribute additional movement dependencies in DC (reflected in the “+1/+2” notation). In RCF, CF increases by +1 in total when such additional dependencies are coded; NF and MF are unchanged.
Source: Adapted from Yang (2026).
Across the corpus, wh-placement reflects the typological contrast between the two lan- guages: Polish-only wh-questions are near-categorical in fronting, Mandarin-only ques- tions are predominantly in situ, and Polish-matrix (code-mixed) questions fall between these extremes (tab. 2). The distribution also changes over time. In Stage 1, Polish-directed wh-questions are largely realized as Polish-matrix clauses with embedded Mandarin wh-material (beyond a small set of formulaic items such as co to?), and wh-placement is variable in this context. In Stages 2-3, fully Polish wh-questions become productive and show target-like fronting, which provides a baseline for evaluating deviations under Polish-matrix code-mixing. In Stage 3, code-mixing in wh-questions becomes rare, but Mandarin questions still show a wh-type asymmetry, so the comparison can be extended beyond mixed clauses.
Tab. 2. Wh-fronting (movement) rates by language context2
Context | Wh-fronting | Wh in situ | Total |
Polish-only | 83 (98.81%) | 1 (1.19%) | 84 |
Polish-matrix (code- -mixed) | 8 (33.3%) | 16 (66.67%) | 24 |
Mandarin-only | 9 (11.39%) | 70 (88.61%) | 79 |
Source: Author’s calculations based on data reported in Yang (2026).
‘What’ questions across stages and language contexts
In Polish-only production, ‘what’ questions are overwhelmingly fronted in the Polish-dom- inant period, establishing a stable target-like baseline. Against this baseline, Polish-ma- trix (code-mixed) ‘what’ questions show a strikingly different placement pattern. When a Mandarin ‘what’ form (shénme) appears in an otherwise Polish frame, the wh-element is retained in situ rather than being fronted. In the present sample, this holds throughout Stage 1–2 for Polish-matrix ‘what’ (Stage 2 alone: 0/11 fronted; pooled Stage 1–2: 0/12 fronted).
Illustrative tokens are given in (1)–(2), where shénme is embedded in a Polish clause while wh-placement aligns with Mandarin:
To shénme? ‘What is this?’
Na nie mówi się shénme ya? ‘What do you call it?’ (ya = part.)
Polish influence is also visible in Mandarin during Polish dominance, but it is limited in the ‘what’ domain. In Stage 2 Mandarin-only ‘what’ questions, wh-fronting occurs in 3/35 tokens (3 fronted vs 32 in situ), as in (3)-(4):
Shénme shì “sweterek”? (2;07.27) ‘What is “sweterek”?’
Shénme shì Mògá (=Mozart)? (2;09.02) ‘What is Moga (=Mozart)?’
The ‘where’ domain patterns differently. In Polish-matrix (code-mixed) ‘where’ questions, Polish-like fronting can occur even when the wh-material is Mandarin (e.g., zài nǎlǐ ‘at where’). This is already attested in Stage 1: the Polish-matrix ‘where’ subset contains 3 fronted vs 0 in situ tokens. In Stage 2, fronting remains frequent in Polish-matrix ‘where’ questions (5 fronted vs 4 in situ, i.e. 5/9 fronted). Examples are given in (5)–(6):
Zài nǎli Heli jest? ‘Where is Heli?’ (Mandarin wh-phrase + Polish copula)
Zài nǎli oparzyłaś się? ‘Where did you get burnt?’ (Chinese wh-phrase + Polish verb)
2 Table 2 reports all wh-questions for which wh-placement could be classified as fronted or in situ; wh-only and subject-wh utterances were excluded (out of 244 wh-questions in the full corpus). The following sections narrow the analysis to ‘what’ and ‘where’.
Polish influence is also visible in Mandarin during Polish dominance in the ‘where’ domain. In Stage 2 Mandarin-only ‘where’ questions, wh-fronting occurs in 25% of all the tokens (3 fronted vs 9 in situ), illustrated in (7)–(9):
Zài nǎlǐ fàng yì diǎndiǎn táng? (2;05.05) ‘Where (should one) put a bit of sugar?’
Zài nǎlǐ zǐsè de? (2;08.17) ‘Where is the purple one?’
Nǎlǐ shì Mòzhā? (2;09.02) ‘Where is Moza (=Mozart)?’
The results show a wh-type asymmetry in Polish-matrix code-mixing: ‘what’ ques- tions favor in-situ placement, whereas ‘where’ questions are more likely to be fronted. Stage 2 provides the most comparable Polish-matrix subset under Polish dominance (tab. 3): ‘what’ shows no fronting (0/11), while ‘where’ shows fronting in 5/9 tokens; a two-sided Fisher exact test confirms an association between wh-type and wh-placement (p = 0.00813). Pooling Stages 1–2 yields the same direction (0/12 for ‘what’ vs 8/12 fronted for ‘where’) and remains significant (p = 0.00135).
A related domain contrast persists in Mandarin when code-mixing becomes rare. In Stage 3, Mandarin-only ‘where’ questions still show fronting (3 fronted vs 0 in situ), whereas Mandarin-only ‘what’ questions are fully in situ (0 fronted vs 4 in situ). This continuation supports treating the wh-type asymmetry as a property of the system beyond the mixed subset.
Tab. 3. Wh-type selectivity in Polish-matrix (code-mixed) questions (Stage 2)
Polish-matrix (code- -mixed) Stage 2 | Wh-fronting | Wh in situ | Total |
‘what’ | 0 | 11 | 11 |
‘where’ | 5 | 4 | 9 |
Source: Adapted from Yang (2026).
This study shows that cross-linguistic influence in bilingual wh-development can be sharply selective within a single construction class. Under comparable dominance con- ditions, the ‘what’ and ‘where’ domains respond differently: Polish-matrix code-mixing favors Mandarin-like in-situ placement for ‘what’, whereas ‘where’ more readily supports Polish-like fronting even with Mandarin wh-material, and the Mandarin data in the later phase preserve the same domain contrast. The implication is straightforward. Domi- nance and input shifts modulate overall activation and availability of patterns, but they do not, on their own, predict which wh-subdomain will be the primary locus of structur- al transfer. This framing is consistent with approaches that treat bilingual outcomes as
shaped by an interaction between language-external factors (dominance, ecology) and language-internal constraints on computation and mapping (Armon-Lotem, Meir, 2018; Meir, Janssen, 2021).
Sections “Why dominance alone cannot account for the wh-type asymmetry”, “A com- plexity-modulated account” and “Relation to previous findings and broader literature” spell out why dominance is insufficient for the observed split, how complexity-based pressures target the right level of explanation, and how the pattern relates to prior wh-transfer findings.
The dominance-oriented analysis developed for this child in Yang (2025) captures the macro-trajectory of interrogative development: periods of increased productivity and the activation of constructions coincide with changes in exposure and use. However, the co/ shénme vs. gdzie/nǎlǐ split requires a more fine-grained explanation because it emerges within the same dominance ecology.
Under early Chinese dominance, Polish-matrix ‘where’ already allows Polish-like fronting with Mandarin wh-material, whereas Polish-matrix ‘what’ remains uniformly in situ; later, when Chinese dominance returns, Polish-like influence does not recede uniformly but persists more in Mandarin ‘where’ than in Mandarin ‘what’.
More generally, this kind of underdetermination is not unique to the present dataset: wh-transfer is not automatic even when dominance and typology appear to favor it, and several longitudinal studies argue that additional constraints are needed to predict when transfer surfaces and in which subdomains it concentrates (Strik, 2012; Park-Johnson, 2017; Qi, Di Biase, 2020). The present study adds a complementary contribution: rather than asking only whether wh-in-situ transfer occurs, it demonstrates how bilingual inter- action can be partitioned within a single domain, concentrating where the target system demands heavier mapping effort.
A complexity-based account is designed to explain selective vulnerability under com- petition between options. Derivational complexity approaches such as the Derivational Complexity Hypothesis (DCH) predict preference for less computationally demanding structures (e.g., fewer movement steps), and have been applied to wh-development and bilingual wh-transfer (Jakubowicz, Strik, 2008; Jakubowicz, 2011; Strik, 2012). This captures the global typological contrast (Mandarin in situ vs Polish fronting, DC=1 vs DC=2~4) but does not explain the within-Polish asymmetry, because co and gdzie are derivationally comparable (both canonically front). As discussed in Yang (2026), DCH can explain why Polish interrogatives are globally more complex than Mandarin, but it cannot explain why co is more difficult/vulnerable than gdzie among Polish wh-fronting questions; that finer split motivates a feature-reassembly perspective.
Feature-reassembly complexity (RCF) supplies that finer grain by treating vulnera- bility as a mapping problem: the burden increases when the learner must stabilize richer feature bundles and less transparent form-meaning correspondences across non-isomor- phic systems (Dai, 2011; Liu, Dai, Li, 2013; Mao, Dai, 2015; Mao, 2020). In the present language pair, the main differentiator is mapping load in the Polish wh-lexeme: gdzie is largely morphologically invariant (MF = 1), whereas co must be realized through multiple case-conditioned alternants (MF = 4 for co/czego/czemu/czym), increasing the transparency cost of mapping between morphosyntactic requirements and surface form. This yields the hierarchy in Table 1 (Mandarin < Polish gdzie < Polish co), which correctly predicts that the Mandarin in-situ option is especially attractive in the higher-mapping-load Pol- ish ‘what’ domain, whereas the Polish ‘where’ domain is more resilient in Polish-matrix code-mixing and can also leave longer-lasting traces on Mandarin later, even under compa- rable dominance conditions. The important point is not that ‘where’ is “easy” in an absolute sense, but that it is less mapping-intensive than ‘what’ in the Polish system, and therefore less likely to trigger a shift toward the Mandarin strategy when the two systems compete.
Interface-based accounts predict selective optionality in bilingual development where syntax interacts with discourse/pragmatics (Hulk, Müller, 2000; Sorace, Serratrice, 2009). The present pattern is compatible with selectivity in general, but it localizes the contrast more narrowly: the crucial split is within wh-questions, where vulnerability tracks item-lev- el morphosyntactic mapping burden rather than wh-fronting as such.
The argument also complements derivational complexity approaches to bilingual transfer. Jakubowicz’s (2011) formulation treats derivational complexity as a condition on transfer: structures that constitute a computational shortcut are more likely to be intro- duced under bilingual interaction. In the present case, the shortcut is not simply to avoid fronting everywhere (since Polish fronting is stable in Polish-only questions once produc- tive), but selectively avoid the most mapping-intensive portion of the Polish wh-system by retaining the Mandarin in-situ configuration in Polish-matrix mixing for ‘what’.
A final control concerns monolingual baselines. Mandarin content questions emerge early and remain predominantly in situ (Li, 1995; Li, Chen, 1998; Fan, 2012). Polish child data likewise show early fronting of both co and gdzie in wh-questions, although productive control of the richer case morphology relevant to the co-domain stabilizes later (Zarębina, 1965; Smoczyńska, 1985). Recent experimental work further illustrates that Polish-speaking children are sensitive to fine-grained properties of wh-questions such as exhaustivity and interpretation (Foryś-Nogala et al., 2017). The baseline therefore does not motivate an age-of-onset explanation for the bilingual ‘what-where’ split. Instead, it supports the present inference: selective vulnerability is best located in the mapping demands imposed by the Polish co-domain under bilingual competition.
This is a single-child longitudinal case study, which gives high developmental resolution but limits generalization. Replication with additional Polish-Mandarin bilingual children
is needed, ideally with different ecological profiles (e.g., different environmental-lan- guage support, as highlighted by Qi and Di Biase, 2020). Experimental work could also probe whether production asymmetries track comprehension/processing costs in wh-de- pendencies. Finally, extending the analysis to additional wh-types (e.g., ‘who’, ‘which’, ‘how’) would test whether mapping burden predicts graded susceptibility across a broader interrogative inventory.
This paper has argued that computational complexity provides an essential ingredient in explaining wh-question development in a Polish-Mandarin bilingual child. Within the same longitudinal study and under the same broad dominance regimes, ‘what’ and ‘where’ questions display systematically different susceptibility to cross-linguistic influ- ence.
The core empirical pattern is an asymmetry: Polish-matrix (code-mixed) ‘what’ ques- tions show strong and stable Chinese influence via wh in-situ, whereas Polish-matrix (code- mixed) ‘where’ questions exhibit earlier and more persistent Polish-like fronting – even when Chinese wh-lexemes are used – and Polish influence remains detectable in Mandarin ‘where’ questions during the later return to Chinese dominance.
These findings support a complexity-modulated view of bilingual development: dominance helps explain macro-level activation and dormancy, but the micro-level dis- tribution of CLI across closely related constructions is shaped by the relative complexity of the target representations and operations. For Polish-Mandarin bilinguals, the higher feature-reassembly burden of the Polish co-domain makes Chinese strategies especially attractive, while the lower complexity of the gdzie-domain facilitates earlier convergence on Polish movement and greater resilience over time.
From an applied perspective, the findings suggest that support for bilingual develop- ment may need to be domain-specific: under comparable exposure conditions, ‘what’ questions can remain more fragile than ‘where’ questions, which resonates with pedago- gical discussions of uneven bilingual language skill and targeted support (Szurek, 2018).
Armon-Lotem S., Meir N. (2018), The nature of exposure and input in early bilingualism, [in:] A. De Houwer, L. Ortega (eds.), The Cambridge handbook of bilingualism, Cambridge: Cambridge University Press, pp. 193–212, https://doi.org/10.1017/9781316831922.011
Benmamoun E., Montrul S., Polinsky M. (2013), Defining an “ideal” heritage speaker: The- oretical and methodological challenges, “Theoretical Linguistics”, vol. 39(3–4), pp. 259–294, https://doi.org/10.1515/tl-2013-0018
Cheng L.L.S. (1997), On the typology of wh-questions, New York: Garland.
Chomsky N. (2001), Derivation by phase, [in:] M. Kenstowicz (ed.), Ken Hale: A life in lan- guage, Cambridge, MA: MIT Press, pp. 1–52, https://doi.org/10.7551/mitpress/4056.003.0004
Cuza A., Pérez-Tattam R. (2016), Grammatical gender selection and phrasal word order in child heritage Spanish: A feature re-assembly approach, “Bilingualism: Language and Cognition”, vol. 19(1), pp. 50–68, https://doi.org/10.1017/S1366728914000893
Dai M. (2011), Er yu xide zhong de tezheng yu tezheng zuzhuang [二语习得中的特征与特征组装], “Yuyan Jiaoxue yu Yanjiu” [语言教学与研究], vol. 4, pp. 91–99.
De Houwer A. (2009), Bilingual first language acquisition, Bristol: Multilingual Matters, https://doi.org/10.21832/9781847691507
Deuchar M., Muntz R. (2003), Factors accounting for code-mixing in an early developing bilingual, [in:] N. Müller (ed.), (In)vulnerable Domains in Multilingualism, Amsterdam: John Benjamins, pp. 161–190, https://doi.org/10.1075/hsm.1.07deu
Fan L. (2012), Ertong zaoqi yuyan zhong yiwen ci de yiwen he fei yiwen yongfa [儿童早期语言中疑问词的疑问和非疑问用法], “Huawen Jiaoxue yu Yanjiu” [华文教学与研究], vol. 1, pp. 85–94.
Foryś-Nogala M., Haman E., Katsos N., Krajewski G., Schulz P. (2017), Syntactic, semantic and pragmatic correlates of the acquisition of exhaustivity in Wh-questions: A study of Polish monolingual children, “Language Acquisition”, vol. 24(1), pp. 27-51, https://doi.org/10.108 0/10489223.2016.1179744
Gębka-Wolak M., Walkiewicz A. (2013), Teaching Polish word order to foreigners: From theoretical assumptions to practical solutions, “Kwartalnik Neofilologiczny”, vol. 60(4),
pp. 397–412.
Goldrick M., Putnam M., Schwarz L. (2016), Coactivation in bilingual grammars: A com- putational account of code mixing, “Bilingualism: Language and Cognition”, vol. 19(5),
pp. 857–876, https://doi.org/10.1017/S1366728915000802
Hu J. (2019), Prominence and locality in grammar: The syntax and semantics of wh-questions and reflexives, London–New York: Routledge.
Huang C.-T.J. (1982), Logical relations in Chinese and the theory of grammar, doctoral dis- sertation, Cambridge, MA: Massachusetts Institute of Technology.
Hulk A., Müller N. (2000), Bilingual first language acquisition at the interface between syntax and pragmatics, “Bilingualism: Language and Cognition”, vol. 3(3), pp. 227–244, https://doi. org/10.1017/S1366728900000353
Jakubowicz C., Strik N. (2008), Scope-marking strategies in the acquisition of long-distance wh-questions in French and Dutch, “Language and Speech”, vol. 51(1–2), pp. 101–132, https:// doi.org/10.1177/00238309080510010701
Jakubowicz C. (2011), Measuring derivational complexity: New evidence from typically developing and SLI learners of L1 French, “Lingua”, vol. 121(3), pp. 339–351, https://doi. org/10.1016/j.lingua.2010.10.006
Kupisch T. (2008), Dominance, mixing and cross-linguistic influence: On their relation in bi- lingual development, [in:] P. Guijarro-Fuentes, M.P. Larrañaga, J. Clibbens (eds.), First Lan- guage Acquisition of Morphology and Syntax: Perspectives across Languages and Learners, Amsterdam: John Benjamins, pp. 209–234, https://doi.org/10.1075/lald.45.09kup
Lanza E. (1997), Language mixing in infant bilingualism: A sociolinguistic perspective, Oxford: Clarendon Press.
Lardiere D. (2008), Feature assembly in second language acquisition, [in:] J.M. Liceras,
H. Zobl, H. Goodluck (eds.), The role of formal features in second language acquisition, Mahwah, NJ: Lawrence Erlbaum Associates, pp. 106–140.
Lardiere D. (2009), Some thoughts on the contrastive analysis of features in second lan- guage acquisition, “Second Language Research”, vol. 25(2), pp. 173–227, https://doi. org/10.1177/0267658308100283
Li Y., Chen Q. (1998), Yuyan de lijie yu fasheng: Ertong wenju xitong de lijie yu fasheng de bijiao yanjiu [语言的理解与发生: 儿童问句系统的理解与发生的比较研究], Wuhan: Huazhong Shifan Daxue Chubanshe.
Li Y. (1995), Ertong yuyan de fazhan [儿童语言的发展], 1st ed., Wuhan: Huazhong Shifan Daxue Chubanshe.
Liu A., Dai M., Li Z. (2013), Tezheng zuzhuang shijiao de Yingyu guanci xide yanjiu [特征组装视角的英语冠词习得研究], “Waiyu Jiaoxue yu Yanjiu” [外语教学与研究], vol. 45(3),
pp. 385–397.
Mao T., Dai M. (2015), Er yu tezheng chongzu jiashuo zhi goujia yu yuanjing pingshi [二语特征重组假说之构架与远景评释], “Zhongnan Daxue Xuebao (Shehui Kexue Ban)” [中南大学学报(社会科学版)], vol. 21(1), pp. 251–256.
Mao T. (2020), Zhongguo Yingyu xuexizhe daoju yulei xide yanjiu [中国英语学习者导句语类习得研究], Beijing: Tsinghua University Press.
Meir N., Janssen B. (2021), Child heritage language development: An interplay between cross-linguistic influence and language-external factors, “Frontiers in Psychology”, vol. 12, article 651730, https://doi.org/10.3389/fpsyg.2021.651730
Meir N., Walters J., Armon-Lotem S. (2017), Bi-directional cross-linguistic influence in bilingual Russian-Hebrew children, “Linguistic Approaches to Bilingualism”, vol. 7(5),
pp. 514–553, https://doi.org/10.1075/lab.15007.mei
Montrul S. (2016), The acquisition of heritage languages, Cambridge: Cambridge University Press, https://doi.org/10.1017/CBO9781139030502
Müller N., Hulk A. (2001), Crosslinguistic influence in bilingual language acquisition: Ita- lian and French as recipient languages, “Bilingualism: Language and Cognition”, vol. 4(1),
pp. 1–21, https://doi.org/10.1017/S1366728901000116
Paradis J. (2011), Individual differences in child English second language acquisition: Com- paring child-internal and child-external factors, “Linguistic Approaches to Bilingualism”, vol. 1(3), pp. 213–237, https://doi.org/10.1075/lab.1.3.01par
Paradis J., Genesee F. (1996), Syntactic acquisition in bilingual children: Autonomous or interdependent?, “Studies in Second Language Acquisition”, vol. 18(1), pp. 1–25, https://doi. org/10.1017/S0272263100014662
Park-Johnson S.K. (2017), Crosslinguistic influence of wh-in-situ questions by Korean-English bilingual children, “International Journal of Bilingualism”, vol. 21(4), pp. 419–432, https:// doi.org/10.1177/1367006916629224
Putnam M., Sánchez L. (2013), What’s so incomplete about incomplete acquisition? A prole- gomenon to modeling heritage language grammars, “Linguistic Approaches to Bilingualism”, vol. 3(4), pp. 478–508, https://doi.org/10.1075/lab.3.4.04put
Qi R., Di Biase B. (2020), The influence of the environmental language (Lε) in Mandarin-
-English bilingual development: The case of transfer in wh-questions, “International Journal of Bilingualism”, vol. 24(4), pp. 691–714, https://doi.org/10.1177/1367006919876716
Slabakova R. (2009), Features or parameters: Which one makes second language acquisition easier, and more interesting to study?, “Second Language Research”, vol. 25(2), pp. 313–324, https://doi.org/10.1177/0267658308100291
Smoczyńska M. (1985), The acquisition of Polish, [in:] D.I. Slobin (ed.), The crosslin- guistic study of language acquisition, vol. 1, Hillsdale, NJ: Lawrence Erlbaum Associates,
pp. 595–686.
Sorace A. (2011), Pinning down the concept of “interface” in bilingualism, „Linguistic Ap- proaches to Bilingualism”, vol. 1(1), pp. 1–33, https://doi.org/10.1075/lab.1.1.01sor
Sorace A. (2012), Pinning down the concept of interface in bilingual development: A reply to peer commentaries, „Linguistic Approaches to Bilingualism”, vol. 2(2), pp. 209–217, https:// doi.org/10.1075/lab.2.2.04sor
Sorace A., Serratrice L. (2009), Internal and external interfaces in bilingual language de- velopment: Beyond structural overlap, “International Journal of Bilingualism”, vol. 13(2),
pp. 195–210, https://doi.org/10.1177/1367006909339810
Strik N. (2009), Derivational contrasts in Dutch and French: Evidence from the acquisition of long-distance wh-questions, “Linguistics in the Netherlands”, vol. 26, pp. 91–102, https:// doi.org/10.1075/avt.26.09str
Strik N. (2012), Wh-questions in child bilingual acquisition of French: Derivational complexity and cross-linguistic influence, “The Canadian Journal of Linguistics / La Revue canadienne de linguistique”, vol. 57(1), pp. 133–151, https://doi.org/10.1353/cjl.2012.0026
Szurek M. (2018), W jaki sposób można kształtować sprawność językową dziecka bilin- gwalnego?, “Językoznawstwo”, vol. 11, pp. 171–182, https://doi.org/10.25312/2391- 5137.11/2017_171-182
Yang A. (2025), Question Acquisition in a Polish-Chinese Bilingual Child: A Longitudinal Account of Dormancy and Activation, “Polish Journal of Critical Realism”, vol. 1(8), pp. 117- 132, https://doi.org/10.5281/zenodo.22129671
Yang A. (2026), Zhōng Bō shuāngyǔ értóng tèshū yíwènjù xídé yánjiū [中波双语儿童特殊疑问句习得研究, Bilingual acquisition of wh-questions by a Polish-Chinese child] [Doctoral dissertation, Beijing Foreign Studies University].
Yip V., Matthews S. (2000), Syntactic transfer in a Cantonese-English bilingual child, “Bilingualism: Language and Cognition”, vol. 3(3), pp. 193–208, https://doi.org/10.1017/ S136672890000033X
Yip V., Matthews S. (2006), Assessing language dominance in bilingual acquisition: A case for mean length utterance differentials, „Language Assessment Quarterly”, vol. 3(2), pp. 97– 116, https://doi.org/10.1207/s15434311laq0302_2
Yip V., Matthews S. (2007), The bilingual child: Early development and language contact, Cambridge: Cambridge University Press, https://doi.org/10.1017/CBO9780511620744
Yuan B. (2017), Computational complexity and L1 transfer: A study of L2 Chinese attitude-
-bearing questions, “Shijie Hanyu Jiaoxue” [世界汉语教学 / Chinese Teaching in the World], vol. 31, pp. 85–104, https://doi.org/10.13724/j.cnki.ctiw.2017.01.006
Zarębina M. (1965), Kształtowanie się systemu językowego dziecka, Wrocław: Zakład Naro- dowy im. Ossolińskich, Wydawnictwo PAN.