Even data, I'm afraid, is often seen as theory-laden. There is some
interpretation going on even when we report things. If you talk about a
stone going through a glass window, you're making assumptions about the
state and nature of the glass before and after the event, talking about
properties of glass containing counterfactuals (unbroken glass is fragile,
meaning that _if_ it were to have a stone thrown at it, it _would_ break.)
You're classifying "glass windows" as transparent sheets made of a
category of compounds based on aggregate. You're assuming your eyes aren't
deceiving you, and the stone isn't just passing through the glass.
Even needles flickering on electronic equipment are theory-laden
observations. We don't just say "oh, the ammeter needle is moving 30
degrees. That's interesting". We say "oh look, there are 4.5 amps moving
thorugh that wire" (excuse my physics). Our observations generally have an
extra layer of interpretation even before we start to Interpret them.