Speech-on-speech perception relies on perceptual and cognitive mechanisms. Differences in voice cues, such as mean fundamental frequency (F0) and vocal-tract length (VTL), and intensity levels (target-to-masker ratio, TMR), help discriminate between target and masker speech and facilitate speech-on-speech perception. Here, using Dutch Child-friendly Coordinate Response Measure (CCRM) sentences with numbers and colors as keywords, we show that masking patterns from a single-talker masker vary depending on the masker type. We used scrambled sentence maskers and intact full sentence maskers, where keywords from the target and masker speech largely overlapped. In addition, the effect of voice cue differences between target and masker speech was assessed for the full sentence condition only, by manipulating F0 and VTL of the masker speech. The scrambled sentence masker showed a weak monotonic masking effect as a function of varying TMR. The full sentence masker showed a non-monotonic masking effect, which was affected by TMR and voice differences. Hence, masking from single-talker speech highly depends on acoustic factors and overlapping keywords. The different patterns of speech perception scores as a function of masker type and TMR has implications on methods commonly used to quantify speech-on-speech perception, such as when determining speech reception thresholds using adaptive procedures.