I recently read a report of new neuroscience research in which researchers are able to predict what a person will recall by analyzing their brainstate. You can read a summary here.
This reminds me of an idea I had a while back for using biofeedback to guide brainstates, in order to improve memory. Here's a hypothetical experiment that illustrates the idea. Show a person a set of photographs, and while they are observing each photo use functional brain imaging to record their brainstate. Later, show them the same photos several more times and make additional recordings of their brainstate, in order to generate a database of brainstates that correspond to their perception of each photo. Next, select a photo secretly (without telling the human subject) and lookup its corresponding recorded brainstates in the database. Then, guide the human subject to generate a brainstate that corresponds to the secretly chosen photo using biofeedback that is tied to their real-time brainstate. For example, provide the human subject with a sound or a computer image that corresponds to their real-time brainstate, and which provides them with positive or negative feedback based on the "distance" from their present brainstate to the desired target brainstate, enabling them to guide their brainstate the correct configuration. After the subject becomes accustomed to using the biofeedback system, apply it to guide them to generate a brainstate that matches or is closely within range of the desired brainstates for the selected photo. Then ask the subject to report which photo they are thinking of. We can measure how well the method works by the accuracy by which the subject reports thinking of the photo we selected originally.
If this process works it could be used someday as a new kind of memory aid. For example, suppose that someday functional brain imaging gets small and portable, or even wearable or implantable, so that everyone has access to their real-time brainstate data. When they want to "remember" something they simply hit the "record" button on their personal brainstate recorder and it measures their brainstate while they are thinking of and/or perceiving what they want to recall. Then they simply give this dataset a label or filename in their personal memory database. Later when they want to recall a specific thing, they just select the label and the system uses biofeedback to guide them back to generating that brainstate, at which point they can then recall whatever it is they were trying to remember.