Ultan Ó Broin's blog post Natural User Interface Gestures. Are They Global? brought to mind a conversation I had many years ago with a Turkish-German court interpreter in Austria. He told me how judges in Vienna's court system kept misinterpreting Turkish defendants' body language. The gesture in question is the Turkish "no", which is expressed by jerking the head back, then bringing it forward; a gesture that looks very similar to a German nod of approval. The interpreter said that in one instance the judge had told him that his interpretation was incorrect, since "clearly the defendant is agreeing" with a statement the defendant had just vehemently denied.
Similarly, when I first moved to Mexico, I would walk away when my co-workers were gesturing for me to follow them because their gesture for coming (palm down, moving palm towards wrist) looked to me like I was being shushed away.
Just as gestures don't always mean the same thing in different cultures, so pictures and colors don't necessarily convey the same thing everywhere. In the West red is usually associated with danger, in Chinese culture it's the lucky color, certain body parts are considered rude in some cultures, ...
I understand how companies attempt to save money on instructions for their products by making them image-only. That way, the product -- such as the assembly-required chair I just bought for my home office -- can be sold anywhere in the world with the exact same package contents, including the instructions. All you need to do is slap a different label on the outside of the box and, presto, there is your Spanish, or Hindi, or Swedish, item. Besides, you don't need to have all these instructions translated into 20, or 50, or 100, different languages.
However, since images may have different meanings in other cultures, this approach may well backfire. I wonder how a pet-care company's all-image instructions would fare in an Arab country where dogs are considered unclean. Would the cute dalmatian licking the properly assembled pet-food station elicit the "oohs" it might in the West, or would it cause buyers in Cairo to return the product?
The moral of the story is, if a company decides to go the all-image route for their instructions (as Ikea does), it can't cut localization out altogether. Rather, localization professionals need to be involved in the design process for these instructions, so that potential pitfalls can be caught early on. No words doesn't mean no translation/localization required.