This research investigated the effect on estimated lower asymptotes of the instructions to Graduate Record Examination (GRE) examinees about how the test would be scored. This effect was assessed for four different verbal item types (analogies, antonyms, sentence completion, and reading comprehension) using a two-way, unweighted means analysis of variance on estimated c-parameters--the probability that an extremely low ability examinee will get the item correct (chance probability). The LOGIST program was used to estimate item and person parameters for the GRE administered under formula-scoring and right-scoring instructions. Sample sizes ranged from 2,900 to 4,800. It was thought that the probability of a correct response by low ability examinees might systematically differ under rights- and formula-score instructions; this did not occur. It was also thought that the mean c-parameters might vary by item type; this did occur. Most notably, despite much evidence that estimated c's are typically considerably less than the reciprocal of the number of choices for the item, one item type--antonyms--was identified for which this is not so. It is recommended that these findings be taken into account when estimating item parameters. Four data tables conclude the document. (Author/TJH)