This study examined the effect of sentence-combining and kernel-identification practice on the syntactic component of sixth graders’ reading comprehension, as measured by a cloze instrument developed by the authors, and by two subtests from the norm-referenced Test of Reading Comprehension (TORC). The experimental group completed eight open sentence-combining exercises, seven kernel-identification exercises, and eight cloze exercises over a 10-week period (two or three exercises per week). The comparison group completed eight cloze exercises during the same period. When covaried by Comprehensive Tests of Basic Skills Total Reading scores and by pretest scores on the cloze instrument, results on the immediate posttest administration of the cloze instrument were significantly (p ˂ .001) in favor of the experimental group; results on a 6-week delayed administration of the cloze instrument approached but did not reach significance in favor of the experimental group (p ˂ .07). There was no significant difference between experimental and comparison groups on the two TORC subtests. Since the readability of the cloze instrument was estimated at eighth grade level (due primarily to the use of longer and more complex T-units), it was inferred that sentence-combining and kernel-identification training enabled the experimental group to comprehend longer, syntactically more complex sentences and to exhibit a tendency toward retention of this ability over a 6-week period.