A study just published in Language Learning (online) has looked again at this issue in an interesting way and, while supporting this main conclusion, that semantic connections can lead to confusion, also might have run afoul of the same mistake in reasoning described above. Here's the abstract:
Semantic Categories and Context in L2 Vocabulary Learning by Patrick Bolger and Gabriela Zapata (DOI: 10.1111/j.1467-9922.2010.00624.x)
This article extends recent findings that presenting semantically related vocabulary simultaneously inhibits learning. It does so by adding story contexts. Participants learned 32 new labels for known concepts from four different semantic categories in stories that were either semantically related (one category per story) or semantically unrelated (four categories per story). They then completed a semantic-categorization task, followed by a stimulus-match verification task in an eye-tracker. Results suggest that there may be a slight learning advantage in the semantically unrelated condition. However, our findings are better interpreted in terms of how learning occurred and how vocabulary was processed afterward. Additionally, our results suggest that contextual support from the stories may have surmounted much of the disadvantage attributed to semantic relatedness.It is an interesting study, especially the use of eye tracking to identify concepts that had become interlinked. One group of participants studied stories with new labels for semantically related concepts. Here's an example from the paper:
An example of a semantically related story
A second group studied unrelated vocabulary. Again, here's an example from the paper:
An example of a semantically unrelated story
In they eye-tracking part of the study, they presented groups of a word and a number of images. For example, the participant sees floop, the new word for dog, together with a picture of a dog, a horse, and a table and is asked to identify which picture matched the word floop. In the group that had studied a bunch of animal names, participants eyes lingered longer over the pictures of the horse than the table and they took longer to choose the dog than the group that had studied semantically unrelated words. In the second group, horse was distracting, but not to the same extent, and even though a table had been in the story, it wasn't particularly distracting. This supports the idea that presenting semantically similar vocabulary in a group is not optimal for vocabulary learning.
The interesting thing was that overall, both groups did very well on the tests and both were distracted by semantically related distractor. Indeed, the difference in distraction afforded by the horse, in the example above, compared to the table was larger than the difference between groups. The authors suggest that "the weakness of the effect could be due to more depth of processing overall for both learning conditions, which led to ceiling effects on the subsequent tasks for both groups."
Putting vocabulary in a story is a well-known mnemonic technique. This, they hypothesize, might have led to overlearning in the training phase, which resulted in a ceiling effect in the testing phases. Personally, I'm not convinced of the truth of this explanation. And I think it goes back to the same misunderstandings that came out of Tulving's research. Putting a list of items in a story, is a good way to remember the group of items. But that's not what the testing phase was testing. It looked at how well the participants could associate a new label with a familiar concept. There are mnemonic techniques, such as the key word technique, that will do this for us, but I'm not sure that the story technique would be particularly useful in this regard. It seems more likely that the training phase simply had enough repetitions that the vocabulary was learned well.
Either way, I'd like to end with the closing paragraph, which I wholly agree with, both in its advice and in its restraint:
"Finally, our study (along with others) also implied that textbook writers and publishers may need to rethink a few assumptions; namely, it may not be such a good idea to group vocabulary by semantic field, as is often done in notional-functional approaches. Thematic grouping (Tinkham, 1997), for which intervocabulary semantic relations are weak but related to some realistic story event (e.g., a day in the life of a high school student), might be a better approach. That said, the field of SLA has been haunted with unequivocal advice that has led to little gain and sometimes loss. We do not wish engage in overconfident advocacy, especially because our story contexts were different in several ways from foreign language textbooks. If textbook writers and publishers are going to consider thematic organization, they should also think about how this may complicate the generation of useful activities; that is, if thematic grouping hinders the creation of useful learning activities for students to engage in, then it may be a poor idea. However, it seems to us that textbooks are too highly driven by semantic grouping and that there needs to be at least some development in the thematic direction. It cannot be too difficult to overcome the challenges, although writers and publishers may need to be more creative than they have been in the past."