Phonological similarity has long been known as one of the most baffling lexical relation. Words that sound similar to each other (i.e. phonological neighbors) can be both friends and foes, depending on the nature of the processing task. In particular, it has been shown that similar-sounding words are competitors in perception, causing slower and less accurate word recognition (Luce and Pisoni, 1998), but they help each other in production tasks, as words with many neighbors are produced faster and more accurately than those with fewer neighbors (Vitevitch, 2002). Gahl et al. (2012) reported that in spontaneous speech, words from dense neighborhoods were more reduced phonetically than those from sparse neighborhoods, suggesting that micro-level phonetic variation in spontaneous production was more heavily influenced by speaker-oriented forces than listener-oriented forces. However, most previous research on phonological neighborhood density was conducted on the English language, with only a few exceptions on Spanish and other languages. Furthermore, the controversy over the Spanish results (Baus et al., 2008; Vitevitch and Stammer, 2006, 2009) brings into question whether neighborhood effects are consistent cross-linguistically. In this paper, we examine another Romance language, French, for the effects of neighborhood structure on phonetic variation. Previous literature has reported inhibitory effects of neighborhood density on French word recognition in (Dufour and Frauenfelder, 2012), though the effect on French word production is less clear (see Sadat et al. 2014 for a recent discussion). Our data came from the CID corpus (Bertrand et al., 2008), which consisted of eight hours of recordings of one-to-one dialogues from 16 speakers (10F). We extracted from the corpus a set of 5,412 tokens of monosyllabic monomorphemic content words between 1 and 5 phonemes long. Similar to Gahl et al.'s study, we constructed two statistical models on the variation of word duration and degree of vowel dispersion, respectively. The word duration model showed that neighborhood density (normalized for word length in phonemes) was a positive predictor for word duration (β = 0.034, t = 2.12, pMCMC = .03), after controlling for a large number of lexical and contextual factors (baseline duration, lexical frequency, phonotactic probability, part of speech, orthographic length, phonemic length, contextual predictability, speech rate, previous mentions, speaker sex, etc.), but neighbor frequency failed to reach significance in the duration model (t = 0.42, pMCMC > .1). That is to say, words with high neighborhood density tended to be produced with longer duration in French. General results of the duration model persisted when only 2- or 3-phoneme word tokens were tested. Similarly, the vowel dispersion model found that high-density words had significantly more dispersed vowels (β = 0.12, t = 3.23, pMCMC = .001), when everything else was controlled (most of the control factors in the duration model plus of vowel type, vowel duration and phonetic context), with neighbor frequency being non-significant (t = 1.51, pMCMC > .1) again. General findings of the vowel dispersion model persisted across a set of alternative models. Taken together, the current modeling results suggest a pattern in French that is opposite of the one observed in English. Words from dense neighborhoods in the French lexicon are hyperarticulated - as opposed to hypoarticulated - compared with words from sparse neighborhoods. Is the discrepancy due to the inhibitory effects of neighborhood density on French word production (as suggested in Sadat et al., 2014)? Or does it suggest language-specific patterns of phonetic variation, given the linguistic differences between English and French (in morphology, prosody, speech rhythm, etc.)? We discuss each possibility and the implications for the theories of lexical activation and phonetic variation.