The first author will be available during the Poster Session A (12:10-2:00 US Eastern Time) at the following link: https://bluejeans.com/4106813791 Here is a link to some supplementary materials (the folder the QR code goes to): https://drive.google.com/open?id=1GUNwRdu0pfz2h1A0uAhwgHSZHcP_gJfb Based on findings that longer dependencies result in increased processing effort, dependency- length minimization (DLM) has been shown to be a significant factor in explaining crosslinguistic tendencies in constituent order linearization (Futrell et al. 2015). However, this work also finds cross-linguistic variation with regard to how “optimal” languages are (cf. Gildea & Temperley 2010). We focus on two already-proposed explanatory factors for this cross-linguistic variation: headedness (discussed by Futrell et al. 2015), and flexibility (Levshina 2019; Namboodiripad 2019), and suggest that processing differences may be one source of a reduced effect of DLM in some flexible languages. We use formal acceptability experiments to compare Spanish, an SVO language in which each of the six logical orders is grammatical and attested (see 1), to previous work on English, Malayalam, Korean, and Avar, the latter three of which are flexible verb-final languages (Namboodiripad 2019; Namboodiripad, Kim, & Kim 2018; Namboodiripad & Zaslansky 2019). Because Spanish is both flexible and canonically verb-medial, this allows us to test whether flexibility attenuates the preference for shorter dependencies (at least, in simple sentences) in non-verb-final languages. 117 participants in Salamanca, Spain rated transitive sentences with animate subjects and inanimate objects on a 1-7 Likert scale. Stimuli were distributed among lists pseudorandomly using a Latin Square. Each participant heard 5 tokens of each condition (recorded by a Spanish speaker from Salamanca), and 60 filler items which ranged in structure; participants heard sentences just once once before being asked to rate (cf. speeded acceptability judgment experiments). We expected SVO would be rated highest, followed by verb-initial orders (as Spanish is PRO-drop), and then verb-final orders. We also expected that, because these sentences were presented without a discourse context, subject-object order would be preferred to object-subject order, though this preference may be slight (Kaan 2001). The results show that SVO was indeed rated highest, fol- lowed VOS, VSO, SOV, OSV, and finally, OVS. A linear mixed effects model was fitted (random effects = RATING & CONDITION; fixed effects = PARTICIPANT & ITEM, scaled to SVO). Model comparisons showed that CONDITION was a significant predictor of RATING (χ2(5)=249; p>0.001). Pairwise t-tests (Bonferroni corrected) showed no significant difference between the verb-final orders (p=1.00), and a marginal difference between verb-initial orders (p=0.078). Concurrent with the acceptability study, pupil dilation was measured for all participants using a free software (https://github.com/jeeliz/jeelizPupillometry) and an affordable, portable, setup (under $200.00, not including the cost of a computer). Preliminary results show that patterns in pupil dilation (a measure of listening effort) roughly matched the patterns found in acceptability; verb-initial and verb-final orders aligned with each other. For verb-medial orders, there is a late increase in pupil dilation in OVS sentences as compared to SVO sentences. Our interpretation is that this late increase in dilation is consistent with speakers realizing that they just heard an OVS sentence, not an implausible SVO sentence. However, these analyses are exploratory at this point. A comparison across languages reveals a striking result: When the verb is in non-canonical position (non-verb-final positions for Avar, Korean, and Malayalam, and non-verb-medial positions for Spanish) argument order is not a significant predictor of acceptability. This holds in every language except English, in which only SVO and OSV are grammatical. These results call into question explanations for cross-linguistic similarities based on headedness. Further, Spanish is more flexible than Korean, despite Korean having case-marking on all arguments; these results support recent corpus-based findings that case-marking does not correlate with flexibility, given a non-categorical approach to crosslinguistic variation (Levshina, submitted).