- Associate Professor, Department of Rehabilitation Medicine
- Institute Scientist, Moss Rehabilitation Research Institute
Language & Learning Laboratory
Moss Rehabilitation Research Institute
50 Township Line Road, Elkins Park, PA 19027
Dr. Erica Middleton directs the Language and Learning Lab, which studies language production, learning and language change, and rehabilitation of acquired language impairment (aphasia). A main focus is to understand how words are mentally represented and produced, both in healthy speakers as well as in people who have aphasia from stroke. A major emphasis in the lab is to delineate how treatments for aphasia can be designed in accord with fundamental principles of human learning to maximize and sustain recovery.
Research Focus Areas
Leveraging distributed practice principles to make aphasia rehabilitation more efficient
Difficulty retrieving words for everyday, familiar objects is common in aphasia, which adversely affects a person’s ability to communicate. People with aphasia may retrieve the wrong word (e.g., semantic error such as “orange” instead of “apple”), or they may not be able to retrieve a word at all when asked to name a familiar object. Our prior work has shown that retrieval practice naming treatment, or practice naming items followed by correct-answer feedback, confers persistent benefits to future naming of those items in aphasia. One line of our work examines how distributed practice principles and the amount of practice affects retrieval practice naming treatment benefit and efficiency in aphasia. For example, in one group study of people with aphasia, errorful naming items were administered for retrieval practice training in multiple sessions. The timing and amount of retrievals per item within and across sessions was manipulated. The study found that increasing the number of retrievals per item within a session conferred inconsistent benefits at the cost of hundreds of additional training trials per participant. In contrast, controlling for the number of retrievals per item, there was a striking advantage from distributing items’ retrievals across rather than within a session on later naming performance; and, this advantage came at the cost of very few additional training trials per participant. That is, treatment benefits are the most robust and efficient if the amount of training per item is increased by adding additional training sessions rather than trials within a session, a benefit that arises from distributing practice over sessions.
Spontaneous speech monitoring and incremental learning in language production
Another line of work examines the effects of lexical learning from spontaneous naming error monitoring in aphasia. A key finding, which we have replicated, is that spontaneously corrected semantic errors (e.g., producing tank for jeep) are associated with greater learning compared to errors that are not detected (a learning-from-monitoring effect). In one study, by providing feedback in the form of the correct target word after a randomly selected subset of trials, we established that detection plays no detectable role in the learning-from-monitoring effect, as there was no learning difference between detected (not corrected) trials followed by feedback versus not-detected trials followed by feedback. We also established the learning-from-monitoring-effect is not attributable to simply hearing/repeating the target word after an error, because self-corrected errors are still associated with robust learning compared to detected (not corrected) errors followed by feedback. These results provided support for a theoretical framework of the role of error-driven learning in exerting persistent change in the lexical access system.
Semantic context and effortful retrieval practice effects in aphasia
One line of work seeks to examine how semantic competition may be manipulated to enhance the benefits from retrieval practice naming treatment in aphasia. A prior study found enhanced retrieval failure with greater semantic competition during training (e.g., practice naming groups of items that are more semantically similar by categorical groups such as shirt, jeans, jacket, etc.), but an opposing benefit at test, in support of an error-driven learning mechanism underlying effortful retrieval practice effects in aphasia. That is, an incremental learning mechanism computes the difference between desired states of activation and observed states of activation of lexical units, with greater difference between the two states leading to greater strengthening between the links from semantics to the target word.
Treating lexical-semantic deficits in aphasia
In addition to problems naming objects, people with aphasia sometimes experience difficulty reliably accessing the meanings of words (lexical-semantic deficit). One line of work examines how retrieval practice principles may impact treatment of lexical-semantic deficits. In a recent study, in individuals with aphasia with lexical-semantic impairment, a receptive form of retrieval practice (pick which of two images matches a word) versus restudy (participant is shown the correct image) conferred equivalent benefit to later comprehension performance. Both training conditions exceeded an untreated control condition. Furthermore, retrieval practice and errorless learning naming treatment (production-based forms of practice) both conferred benefits to comprehension compared to untreated. This is evidence of transfer of performance from one modality to another. Additional benefits from retrieval practice naming treatment over errorless learning were observed in the form of more durable benefits to comprehension and greater generalization, indicative of refinement of semantic representations compared to errorless learning.