Raddoppiamento fonosintattico

A few months ago a fellow student of mine asked me "Italian has some kind of grammatical gemination, right?" - I first panicked and thought there was some inflectional category expressed merely through gemination (with gemination as its only exponent) which the Italian Insegnamento Institutions forgot to tell us about. What he probably meant was the so-called raddoppiamento fonosintattico which has nothing to do with inflection or any syntactical categories but owes the sintattico part of its denomination to the fact that it occurs between word boundaries. As Irene Amato of Leipzig University puts it:

The initial consonant of WORD2 in the string WORD1-WORD2 is lengthened if
WORD1 is:

(1) an item of a closed lexical class: 

/'kome/ /'va/! ['ko:me'v:a] ‘how are you?’

(2) stressed on the final syllable: 

/t͡ʃi't:a/ /'kara/! [t͡ʃi't:ak:a:ra] ‘dear town’

Lexical R[addoppiamento] F[onosintattico] (1) is due to a final consonant in the historically earlier form of WORD1(Lat. ad > It. a, Lat. quomodoet > It. come). Stress-driven RF (2) is a phonologically predictable stress triggered
gemination. Importantly, RF-geminates are only 50% longer than singletons, in contrast
to inherent geminates, which are 200% longer (as in ['pas:i] ‘steps’ vs. ['ba:zi] ‘bases’.

The Tuscan lenition or Gorgia (Throat), i.e. the hoha hola hon la hanuccia horta horta thing they've going on down there, may only occur when Raddoppiamento fonosintattico doesn't occur. So you get a distribution like this:

(3)
a. /la/ /'kasa/ ‘the house’!  [la'xa:za]
b. /a/ /'kasa/ ‘at home’! [a'k:a:za]
c. /in/ /'kasa/ ‘in (the) house’! [iŋ'ka:za]

 For more details and an elegant derivation check out Irene's paper.