I know the probes are specific, but as in PCR, sometimes you might be annealing with something else, specially if you designed the probe yourself and its specificity hasn't been validated.
may I ask how large your probe is (how many nucleotides)? When we design our RNA probes for in situ hybridization we usually use probes anywhere between 500 -1500bp and we tend to include parts of the 5' or the 3' UTR of the particular gene to increase specificity.
You gave the example of unspecific binding of primers, but you have to see, that primers are only a few nucleotides and the probability of them binding elsewhere is much larger. Also, you could blast the sequence of your probe against the genome of your model organism to see if it could potentially bind elsewhere.
Use appropriate controls. use a tissue where your gene is not expressed, also use unrelated probe (generated by scrambling) of similar size on your target tissue. There will always be some non specific staining. Optimize the condition (conc, time etc.) in such a way that there is a huge difference between your positive and negative controls. All the Best.
Simone, I'm at the moment just designing probes and thinking of the best way to address my questions through ISH, and that reply of yours actually helped me a lot along with Chellakkan's. I need to find some suitable tissues for positive and negative controls for my probes.