This paper investigates the acquisition of two types of comparative adjectives in English (synthetic comparatives with -er and periphrastic comparatives with more). Graziano-King (1999) observed that English-speaking children aged four tended to prefer periphrasitic comparatives when asked which comparative form of gradable adjectives sounded better, and concluded that (1) periphrasitic comparatives are the default option: more is inserted as a last resort, to save a stranded comparative morpheme ('CMPR'), and (2) once synthetic comparatives are listed in the lexicon, by virtue of positive evidence, they block periphrastic comparatives. To check whether English-speaking children prefer periphrastic comparatives in their spontaneous speech as well, spontaneous-speech data from four English-speaking children in the CHILDES database (Abe, Adam, Sarah, and Naomi) were analyzed. It was found that (1) when comparative forms were produced, they were always synthetic comparatives (not a single periphrastic comparative was produced), and (2) the children did not use the synthetic comparatives together with than-clauses until relatively late. The results suggest that the lexicon of English-speaking children around age four lacks CMPR, and that the synthetic comparatives children produce are not true comparatives.