This talk concerns the question of the competition and interplay between linguistic constraints, phonological, morphological and semantic, in word formation processes. The aim of the present study is to demonstrate that phonological size properties are at work in French word formation and can override morphological and semantic properties of the same rule (about phonological size properties, see the prosodic constraints in McCarthy & Prince 1993 and Prince & Smolensky 1993). For this purpose, we have adopted a two-pronged approach combining corpus data and neurophysiological data on the processing of French Verb-Noun compounds. To date, the few studies that have examined the role of phonological size properties in French word formation were mainly concerned with the substitution of affixes (e.g., Plénat & Roché 2003, Plénat 2009, Lignon 2013). In the present study, we investigate the nominal Verb-Noun (VN) compounding rule in French (ouvreV-boîteN 'can-opener', essuieV-glaceN 'windshield-wiper', gardeV-côteN 'coastguard'), which is considered by the language processing system as a morphological rather than a syntactic process (Scalise 1992, Corbin 1992, Fradin 2009, Villoing 2012a), and therefore is part of lexemic morphology (cf. Anderson 1992, Aronoff 1994, Fradin 2003). Linguistic studies based on Villoing’s corpus on French VN compounds (Villoing 2002) have shown that VN compounding is constrained by phonological size restrictions that have some consequences on the selection of compound constituents, the verbal and nominal bases (Villoing 2009, 2012b; Fradin 2009). • Phonological properties: the size constraint The examination of the syllabic format of VN compounds in French and the syllabic computation of the component verbs and nouns reveals size constraints on the formation process: (i) the VN compounds have an optimal size of two or three syllables; (ii) unlike French verbs in general (Lexique 3, New B., Pallier C., Ferrand L., Matos R., 2001), the verb of the compound has to be monosyllabic (80% of verbs in the corpus of Villoing); (ii) a monosyllabic noun is preferred (60% of the corpus of Villoing); longer nouns may occur only when the verbal basis is short (fume-cigarette 'cigarette holder', porte-hélicoptères 'helicopter carrier'). • Consequence 1: competition with morphological constraints The impact of the phonological size constraints on the compounding process leads to the exclusion of certain verb types whose stem would be too long, despite their appropriate semantic properties: (i) verbs formed through affixation (no verbs with the suffixes -ifier,-iser or -oyer; scarcity of prefixal verbal bases (e.g., adoucir 'soften', ennoblir 'ennoble', décoloniser 'decolonize'); (ii) verbs of the 2nd group (noircir 'blacken', grossir 'enlarge', abolir 'abolish'), for which the selected stem (the third verbal stem in the framework of Bonami & Boyé (2003), that is, the present singular indicative) would be at least disyllabic (noirci-, grossi-, aboli-). • Consequence 2: competition with semantic constraints These size constraints also impose a selection of nouns that do not prototypically fulfil the expected semantic requirements. For example, the N purée ‘mashed potatoes’ in presse-purée 'masher' is semantically inappropriate because it fills the semantic role of result and not the expected role of patient (like the majority of N components in such compounds e.g., ouvre-boîte 'can-opener'. cf. Villoing 2009). In this case, purée replaces the semantically appropriate noun, pomme-de-terre 'potato', which is clearly too long for this construction. In order to study the psychological reality of the linguistic hypothesis that French VN have an optimal size of syllabic format (i.e., two or three syllables), a psycholinguistic experiment is designed. Our aim is to generate empirical data that would allow us to discuss the functional architecture of a model of French VN compound processing (for a model of this kind, see Isel, Gunter, and Friederici, 2003, and revisited by Isel, 2010). This model attempts to account for the processing of spoken German NN compounds on the basis of both behavioral reaction time and neurophysiological data. Furthermore, a previous ERP study examined the role of the lexical frequency of VN constituents during the processing of VN selected from the corpus of Villoing (Rosenfeld, Isel, & Fradin, 2014). Our theoretical hypothesis is that if native speakers of French have a mental representation of an optimal syllabic format for both the VN and their constituents, then the processing of French VN corresponding to this format should be facilitated. The experiment uses an oddball paradigm (deviant stimulus; Näätänen, Gaillard, & Mantysalo, 1978). Behavioral (reaction times and percentages of error) and electroencephalographical (event-related brain potentials, ERPs) data will be collected. • MMN oddball experiment The aim of the experiment is to study whether the electrical responses of the brain (here, the MisMatch Negativity, MMN) of native listeners of French is sensitive to the conditions of syllabic format optimality in French VN. The MMN is considered in the literature as a marker of detection of change. Its rationale relies on the habituation paradigm. If a system detects changes caused by a stimulus, which differs on one property with respect to a series of preceding stimuli, then the system reacts. Here, we study the extent to which the brain is able to detect in real time syllabic structure changes. Concretely, we will present a series of VN having the same syllabic format (standard stimuli), unexpectedly interrupted by a VN showing another syllabic format (deviant stimulus). If the property of syllabic size is considered as a relevant constraint for the language processing system, then changes in syllabic format (deviant stimulus), all things being equal otherwise, should give rise to a MMN variation. Moreover, the MMN should be larger as the syllabic format moves away from the optimal format. Furthermore, we predict that the size of the MMN should also vary as a function of the syllabic format of the internal constituents of the VN. Critically, the largest MMN should be observed for deviant stimuli, which do not show the optimal syllabic format in French VN. This is the case for VN comprising four or more syllables. Among these VN, the MMN should be larger for VN with a disyllabic verb (2-2) than for VN with a monosyllabic one (1-3). In contrast, deviant stimuli that meet the conditions of optimality in French compounds should be associated with the lowest MMN response. It should be the case for disyllabic VN comprising a monosyllabic V and a monosyllabic noun (1-1) as well as for the VN with a monosyllabic V and a disyllabic noun (1-2). However, MMN should be larger for 1-2 than for 1-1. Interestingly, we expect that for the trisyllabic VN, the deviant with the stronger impact should be the one with the less optimal syllabic structure, namely VN with a disyllabic verb followed by a monosyllabic noun (2-1). Consequently we predict that the MMN should be larger for 2-1 than for 1-2.