A fundamental property of (novel) denominal verbs is that only part of their meaning is overtly spelled-out, as it becomes clear when considering the unexpressed (verbal and prepositional) structure postulated in multiple theories (See Tables 1 and 3 in Section 2.1, for instance). The meaning of denominals is compressed, implicit rather than explicit, i.e., (to) cherry means much more than cherry (it could mean ‘to eat cherries’, ‘to pick cherries’, ‘to become like a cherry’, a.m.o.). Children have to figure out the interpretation in the absence of an explicit expression of the underlying meaning of the verb. If children have no difficulties decompressing implicit structures, then they should treat explicit and implicit structures in a similar fashion. However, recent work suggests that children might have difficulties with meaning compression (see Martin, Nie, Alexiadou & Guasti 2022; Guasti, Alexiadou & Sauerland 2023). To shed light on this important point, we assume the Meaning First Approach (Sauerland & Alexiadou 2020), a recently developed generative conceptual framework that relies on the idea that language is compression of thought. According to this approach, conceptual representations are built out of innate conceptual primitives, and they are compressed (linearized and articulated) into various linguistic forms based on various factors (linguistic rules, economy, context, a.o.). This model of grammar involves a Generator component and a Compressor component (see Guasti, Alexiadou & Sauerland 2023). The Generator-which is located in the Thought-system and is language-independent- is the structure-building engine that generates complex conceptual representations (CRs) out of primitives (such as causation, agency, conjunction, negation). The Compressor is the morpho-syntactic linguistic engine responsible for linearization, i.e., deciding which meaning components get articulated by which lexical items and in what order. Compression of meaning is tightly related to the likelihood of being understood by other interlocutors: thus, speakers are less likely to compress if this leads to difficulties for the interlocutor. In the case of children, the Meaning First Approach predicts that, if a concept occurs in a conceptual representation, children will realize it lexically (Child Language Undercompression Hypothesis, see Guasti, Alexiadou & Sauerland 2023). This seems to be the case indeed. While adults generally prefer compressing conceptual structures, we find that children seem to opt for a one-to-one correspondence between concepts and forms (Slobin 1973), making undercompression errors, i.e., production errors where they tend to express the underlying conceptual structure. Support for this comes from Martin et al. (2022), who conducted a corpus study on causatives in child French and found out that, unlike adults, French children younger than 4;6 express the implicit causative component of transitive verbs through the verb faire 10% of the time:

Interestingly, causative faire occurs mostly with causative verbs but it is very rarely used with verbs that lack causative semantics such as manger ‘eat’ or pleurer ‘cry’. This shows that children’s production errors reflect an initial tendency to undercompress/fully make explicit the meaning of causative transitives and to express a concord pattern. This is also reflected in children’s use of double negation to express a negative meaning (She didn’t buy nothing- see Thornton, Notley, Moscati & Crain 2016) or of double comparatives to express one single comparison (plus mieux ‘more better’ in French). Interestingly, the tendency towards concord manifests not only in production but in comprehension as well, where children interpret meaningful causative morphemes as semantically superfluous (in particular, Japanese children interpret ‘The monkey had (someone) open it’ as ‘The monkey opened it’- see Yamakoshi et al. 2018) or they interpret double negation as negative concord (Nicolae & Yatsushiro 2020). Other examples of undercompression errors produced by children (see Guasti, Alexiadou & Sauerland 2023) involve dependencies (children tend to lexicalize the unpronounced trace a DP is associated with, as in the one that he lifted it or the song about the airplane that we’re riding in an airplane (3;9)) and antonymic concepts (children tend to produce both the positive and the negative members of a set together, as in mit ohne Butter ‘with without butter’).

Based on such findings from production, we hypothesize that children should have an easier time with transparent mappings of concepts and forms in comprehension as well. We thus put forth the Compressed Meaning is Hard Hypothesis:

Children initially have more ease with explicit meaning.

Our hypothesis is in line with previous proposals about acquisition, which argue that children have an initial bias towards transparency (Slobin 1973; van Hout 1998). As discussed by Guasti, Alexiadou & Sauerland (2023), the Meaning First Approach assumes that children have to map conceptual representations to language and, consequently, predicts that they should have more ease with transparent one-to-one mappings of concepts and forms, which present the advantage of avoiding ambiguity, than with covert and/or conflated ones (many-to-one). This leads to the prediction that children should have more difficulty interpreting SAD verbs, which are compressed forms, than explicit structures which separately spell out the event(s) or state(s) involved and the corresponding nouns/nominal roots. Thus, children should have more ease with ‘to eat cherries’ and with ‘to become like a cherry’ than with ‘to cherry’, and they should also have more ease with ‘to catch a fox’ and ‘to behave like a fox’ than with ‘to fox’ in the corresponding pragmatic contexts.

In addition, we can also make some predictions regarding the decompressed meanings children prefer to ascribe to denominals: children may differ from adults in their decompression preferences, opting for simplicity over plausibility or world knowledge. A cireşi ‘to cherry’ allows a wide range of possible uncompressed meanings: both literal ones (e.g., ‘to eat cherries’) and figurative ones (e.g., ‘to become (red) like a cherry’, i.e., ‘to blush’). In the absence of any disambiguating context, children might observe a Principle of Simplicity in Decompression such as (9), and, consequently, they might prefer the simpler, literal meanings (doing an activity that involves actual cherries) over the more complex ones which involve both a more complex verbal component (e.g., the change-of-state meaning ‘come to be’) and a more complex noun-like component (like a cherry).

When a novel word may have multiple meanings, children decompress to the simpler meaning by default.

Simplicity may be understood in terms of structural semantic meaning. Literal meanings result from merging V with a noun, while figurative meanings result from merging V with a LIKE N component. This component may be represented as a ‘LIKE N’ structure or it may be considered a root (Table 2) (Kiparsky 1997) or even an OBJECT TYPE N (Bleotu & Bloem 2020; 2021), which acquires a manner/property reading in the context. Children are exposed to many nouns, and they can map (novel) nouns to objects/entities from early on (Echols & Marti 2004; Fennell 2006; Waxman & Booth 2001, a.o.). In contrast, n-like similarity is more semantically complex than noun meaning. Moreover, there may be an additional difference between literal and figurative interpretations of denominal verbs in terms of the light verbs occurring in their make-up: literal readings of denominals may be argued to involve the light verb DO (like eating, picking, among others)-which we shall refer to as DO-type verbs- these are prototypical events, and, from an aspectual viewpoint, they are also simpler than changes of state, which may be argued to compose figurative interpretations (e.g., BECOME like a cherry= ‘COME TO BE’). Given these considerations, we may expect children to generally prefer literal over figurative interpretations of denominals in the absence of context.

Associating ambiguous words with simpler meanings would imply that children observe a hierarchy in their mapping preferences: they initially opt for One-to-One mappings (Slobin 1973; Clark 1987), and, in case such mappings are not possible, their next preference is to map a form to as few concepts as possible.

In the presence of a supportive context, however, children may decompress the adequate meanings if their sensitivity to context is strong (see Section 2.4.3). Interestingly, while early theories of polysemy argue for an innate bias for associating a word with a single category and a single meaning (Markman 1989), more recent approaches suggest that polysemy may be widespread and available in child language (Ramiro, Srinivasan, Malt, & Xu 2018; Srinivasan & Rabagliati 2015; 2021). 2-year-olds can recognize multiple meanings of polysemous words, as shown by a recent eye-tracking study (Floyd, Goldberg & Lew-Williams 2020). Moreover, 4–7-year-olds can learn polysemous words, relying on the relation between the different meanings (Srinivasan, Berner & Rabagliati 2019), in particular, the relation of similarity (Floyd & Goldberg 2020). Young children can also interpret nonce denominals in a ‘figurative’ way, focusing on the manner of the activity (Lippeveld & Oshima-Takane 2011; 2013; 2015; 2020) in a context involving direct visual contact with the entities and actions at issue. In the presence of a supportive context, 5-year-old children might thus be able to interpret to cherry as meaning ‘to become (red) like a cherry’, not just as ‘to eat/pick cherries’. If, however, meaning extension in terms of similarity is more challenging (see Section 2.4.2) and children’s preference for simpler meanings is stronger, they might still prefer simpler meanings, i.e., literal meanings- which would not be infelicitous in the context.

According to the Meaning First Approach and the Compressed Meaning is Hard hypothesis, explicit meaning should thus be easier than implicit meaning. Note that, while implicitness may occur in the absence of compression in various situations, for instance, with implicit arguments (see Bhatt & Pancheva (2017) for an overview3), in the particular case of denominal verbs, compression and implicitness go together. Compressing multiple meaning components into a denominal results in those meaning components becoming implicit while decompressing them makes them explicit. Children should have less difficulty making sense of explicit meanings. This holds for literal meanings, but it might also hold for figurative meanings (‘to become like a cherry’, ‘to behave like a fox’). While the explicit paraphrases of figurative meanings of SAD verbs are still open to several interpretations (e.g., ‘to be sly’, ‘to steal’, ‘to lie’ for to fox), these interpretations are fewer than all the possible interpretations for the corresponding implicit denominal, which include both figurative and literal interpretations. However, figurative meanings might generally be more challenging for children, whether explicit or implicit. If the comparison like a cherry/a fox is provided, children would be made aware that the meaning is figurative. However, they would still need to rely on additional world knowledge (involving similarities between objects) to figure out in what respect the meaning is figurative (e.g., size, colour, shape). Consequently, although the explicit figurative meaning is easier than the implicit figurative meaning, it might still be harder than the literal meaning.

Further insight into the matter comes from cognitive accounts of various types of figurative meanings.

From a cognitive perspective, figurative interpretations of SAD verbs should be more problematic than literal ones. To derive an adult-like figurative interpretation, children should (a) understand that the meaning is figurative and (b) understand how the meaning is figurative, i.e., in principle, children could understand that a LIKE component is involved, but still not be able to identify the relevant properties underlying the comparison (we owe this important remark to one of our anonymous reviewers). Children are known to have difficulties with figurative meanings between the ages of 3–5, and it has been shown that there is a developmental path in metaphor comprehension (Billow 1975; Gentner 1988; Johnson & Pascual-Leone 1989; Nall 1983), with older children performing better than younger ones. This is because figurative meanings are cognitively more complex than literal meanings: understanding figurative meanings involves identifying the similarities between objects, establishing relations between those similarities, and integrating them into a new concept (Paivio 1979; Lakoff & Johnson 1980; Recanati 2004; Rubio-Fernández 2007). Children also have more limited world knowledge than adults, therefore, a more limited capacity to identify prototypical similarities between objects. While children may be aware of other types of similarities, requiring less cultural knowledge, children may lack the ability to identify stereotypical similarities. Consequently, we would expect them to experience difficulties with figurative readings of SAD verbs.

An important question is whether children do better with certain types of figurative meanings than with others (Song 2020). The existence of a possible difference between implicit figurative meaning (such as metaphor) and explicit figurative meaning (such as simile) has been a matter of debate. According to the traditional comparison view (Miller 1974; Miller & Johnson-Laird 1976; Miller & Glucksberg 1988; Ortony 1979; 1993; Gentner & Bowdle 2008; Glucksberg, Newsome & Goldvarg 2001), going back to Aristotle, similes are explicit comparisons, whereas metaphors are implicit comparisons. Nevertheless, they essentially represent the same figure of speech asserting the analogy between an object X and an object Y. Metaphors are simply a form of conventionalized comparison. Figuring out the meaning of a metaphor or a simile involves (i) first deriving the literal meaning, (ii) then assessing the interpretability of that meaning, (iii) deriving the non-literal meaning if the literal meaning does not work. Such accounts use data from simile interpretation as evidence for general figurative meaning interpretation (e.g., Gentner 1977; Vosniadou & Ortony 1983). They predict that children should treat implicit figurative meanings and explicit figurative meanings similarly. Thus, on the assumption that denominal verbs involve metaphors (Štekauer 2005; 2006), children should understand Mary cherried when John told her she is beautiful with the same ease as Mary became like a cherry when John told her she is beautiful.

According to accounts that treat similes as different from metaphors, metaphors represent class-inclusion categories: a statement such as x is a y should be understood as ‘x is a member of the category in which y is a prototype’ (Glucksberg & Keysar 1990). In contrast, similes should be understood as comparison statements (Glucksberg & Haught 2006). This is the perspective adopted by Relevance Theory, according to which, while simile comprehension involves comparing the properties of two literal concepts, metaphor comprehension involves constructing an ad-hoc concept that fits the context, thus leading to broadening or narrowing the meaning (Carston 2002; Sperber & Wilson 2008). Chiappe & Kennedy (2001) and Roncero et al. (2006) even argued that metaphors express a stronger meaning than similes since X being a cherry involves more similarities between X and the object cherry than X being like a cherry. The idea of a difference between similes and metaphors has been supported by eye-tracking-while-reading studies indicating longer reading times and higher rate of regressions for metaphors than for similes in adults (Ashby, Roncero, de Almeida, & Agauas 2017), as well as by several child language studies showing that similes are easier to understand than metaphors (e.g., Reynolds & Ortony 1980; Seidenberg & Bernstein 1986; Siltanen 1990). Such accounts predict that children should handle similes such as become like a cherry in Mary became like a cherry when John told her she is beautiful more easily than the (to) cherry, which carries an implicit figurative meaning. This prediction is also partly supported by a previous study by Özçalışkan, Goldin-Meadow, Gentner, & Mylander (2009), who found that, after acquiring the word like, hearing children shift from primarily expressing global similarity (cat/tiger) to primarily expressing single-property similarity (the crayon is brown like my hair), unlike deaf children. Consequently, the authors argued that the explicit marker like might foster similarity comparison.

Interestingly, the predictions made by theories that postulate the non-equivalence of metaphors and similes (Mary is a cherry versus Mary is like a cherry) are somewhat similar to the predictions made by the Meaning First Approach about uncompressed figurative meaning being easier than compressed figurative meaning.

Theories that consider metaphors different from similes (Mary is a cherry vs. Mary is LIKE a cherry) predict that similes should be easier for children than metaphors. Mary is a cherry (or Mary cherried) does not tell participants directly what type of meaning they should embrace (literal/figurative), but this is something they should infer by relying on the knowledge that humans cannot be cherries, they can only do various things with cherries (e.g., eat, pick) or be like cherries in a certain sense. In contrast, the presence of the explicit marker like in the simile Mary is like a cherry makes figurative readings directly available, removing literal interpretations from the array of possible interpretations. On the other hand, theories that treat metaphors on a par with similes, assuming that metaphors also involve a LIKE component, predict that children should treat Mary is a cherry and Mary is like a cherry alike.

As far as the Meaning First Approach is concerned, it assumes that transparent structures (with a one-to-one mapping of meaning to forms) should be easier to acquire than non-transparent structures. If non-transparent figurative denominals involve an implicit LIKE component, then explicit similes using like should be easier for children, given that the LIKE component is directly accessible in become like a cherry and not something to be inferred. If, on the other hand, we assume that, in non-transparent figurative denominals, a light verb simply combines with a root (Kiparsky 1997), then children should still perform better with explicit figurative structures, since, for implicit figurative denominals, they would have to figure out the verb in the underlying structure, whereas for explicit figurative structures, the verb is directly accessible.

Another possibility is that, although, in the absence of any context, children might handle explicit meanings better than implicit meanings, in the presence of supporting context, children actually handle implicit meanings with just as much ease as explicit meanings. Previous research has shown that context plays a very important role in metaphor understanding (Vosniadou 1989; Cameron, 1996; 2003; Özçalişkan 2007; Waggoner, Palermo & Kirsh 2017). In the absence of any context, 4-year-olds have difficulties understanding motion metaphors, for instance. However, when sufficient context is provided, the difficulties fade away. If children are extremely sensitive to context, figurative meanings should pose no extra challenges compared to literal meanings. Thus, the Simplicity in Decompression Principle might be overridden by the Plausibility in Decompression Principle:

When a novel word may have multiple meanings, decompress to the most plausible meaning, given context and world knowledge.

While we expect adults to show sensitivity to context and world knowledge, it is unclear whether children show the same level of sensitivity.

For convenience, Table 6 lists our general expectations about children’s handling of SAD verbs according to the theories discussed. Interestingly, while all the accounts presented below make predictions about figurative meanings, only the Meaning First Approach makes a prediction about literal meanings, i.e., that implicit literal meanings should be harder than explicit literal meanings. It bears mentioning that the approaches in Table 6 may combine in mutually non-exclusive ways if participants compute meaning both structurally and cognitively– we come back to this matter in the discussion.

We conducted a Contextual Denominal Task, a picture selection task where children were exposed to critical sentences containing SAD verbs like a cireşi ‘to cherry’ embedded in a certain pragmatic context, creating a bias either for literal or figurative interpretations. We wanted to see whether Romanian children handle figurative interpretations better when there is supporting pragmatic context.

In another experiment, the Contextual Explicit Denominal Paraphrase Task, we expose Romanian children and adults to explicit paraphrases of denominal verbs in literal- or figurative-biased contexts. We specifically investigate if children find it easier to access figurative readings in figurative contexts when the internal structure of SAD verbs is explicitly spelled out (e.g., to become like a cherry instead of to cherry, to act like a fox instead of to fox).

Previous research on light verbs in child language shows conflicting evidence: some studies show that children prefer using single non-light verbs (e,g., “A kissed B”) over paraphrases with (light) verbs (e.g., “A gave B a kiss”) (Barner 2001; Goldberg 2013; Oshima-Takane et al. 2001; He & Wittenberg 2020), while others argue that children (with and without SLI) tend to overuse light verb constructions (Kambanaros & Grohmann 2013). Paraphrases have more arguments than semantic roles (Wittenberg 2018; He & Wittenberg 2020), but they have the advantage of semantic transparency. Thus, we might expect children to understand paraphrases of SAD verbs better than SAD verbs, given that they make the underlying meaning more explicit.