Abstract Variation in signals often encodes additional information beyond that provided by the main signal. Many species communicate degree of threat using such signal variation. However, multiple signal parameters often covary with threat level, and it is unclear whether receivers are using variation along one or more parameters when assessing threat level. White-throated magpie-jays (Calocitta formosa) vary several call parameters when mobbing, and here I report on an experiment testing whether such variation contains information used by conspecific receivers, and which parameters are salient to receivers. I collected data on natural mobbing sequences in two threat contexts and found that mobbing calls were highly variable in both length and inter-call interval, but on average, both parameters were shorter in high threat contexts, while call frequency did not differ. To determine how jays respond to such between-context variation, I conducted a playback experiment in which call length and inter-call interval were independently varied in a factorial design. This is one of the first studies to my knowledge to deconstruct and independently test covarying signal parameters in an anti-predator signaling system. Magpie-jays responded more often to treatments with either short call lengths or short inter-call intervals, in concordance with the natural calling data. However, the two parameters did not appear to act independently; responses were not additive, and short calls at short inter-call intervals elicited no stronger of a response than short calls at long intervals or long calls at short intervals. Receivers can use either parameter when assessing threat level. These data demonstrate that mobbing can signal threat level, that call length and call rate are redundant, and that receivers need not use both to determine whether to approach.