Modern Hebrew (MH) spirantization is a variable phenomenon with many exceptions. Adam (2002) claims that the variation is driven by these exceptions and concludes that spirantization is changing, yielding what is currently a variable grammar, with equal weight given to expected and variant forms. However, it is not known what the current stage of the grammar looks like. This paper reports the results of an acceptability rating task showing that the expected form is still rated as more acceptable than variant forms, and which variant surfaces (stop or fricative) depends on its underlying root position. MH has three stop/fricative pairs participating in spirantization – [p]/[f], [b]/[v], and [k]/[χ] – with fricatives surfacing in post-vocalic position and stops occurring elsewhere, as illustrated in the verbal paradigms in (1). (1) MH Spirantization Root Infinitive~Uninflected Gloss [p]~[f] /pgʃ/ [lifgoʃ]~[pagaʃ] ‘meet’ [b]~[v] /sbl/ [lisbol]~[saval] ‘suffer’ [k]~[χ] /ʃkv/ [liʃkav]~[ʃaχav] ‘lay down’ Spirantization in MH is complicated by the presence of both exceptions and variation. Exceptions are cases in which [p], [b], and [k] do not alternate with their fricative counterparts and vice versa, such that the stops may surface in post-vocalic position and the fricatives in non-post-vocalic position, as in (2). (2) Exceptions to spirantization(underlined) Root Infinitive~Uninflected Gloss /k/(< *q) /krʔ/ [likro]~[kara](*[liχro]) ‘read’ /v/(< *w) /vtr/ [levater]~[viter](*[biter]) ‘give up’ There is also variation in MH spirantization (Adam 2002, Temkin Martínez 2008), where segments that normally conform to the distribution in (1) surface as stops where fricatives are expected or as fricatives where stops are predicted. (3) Variation in spirantization Root Expected/Variant Gloss /spr/ lispor/lisfor ‘count’ /kbr/ jikbor/jikvor ‘bury’ /ksh/ jeχase/jekase ‘cover’ In the experiment described in this paper, 74 participants rated the naturalness of words pronounced with either the expected or variant forms of alternating and exceptional segments. Each participant was presented with four tokens for each of 20 roots – the expected and variant forms of two conjugations. Tokens were randomized and therefore rated independently from others stemming from the same root. Preliminary results show that, for the alternating roots, only 4 of the 16 possible rating patterns are attested, as in (4). (4) Root patterns for alternating segments a. both expected forms rated natural (highest ratings) b. both expected forms and the variant with a fricative rated natural c. both expected forms and the variant with a stop rated natural d. all four forms rated natural These results clarify the current patterns of variation in MH, and demonstrate that the expected alternating forms are still preferred over variant forms. Further analysis indicates that the acceptability of variant forms (in patterns b and c) is significantly correlated with the root position for the alternating segment, such that there is a preference for stops when the segment is in root-initial position and a preference for fricatives in root-medial position.