The 'Sentence Superiority Effect' is due to guessing

Snell and Grainger (2017) reported a 'Sentence Superiority Effect' (SSE) in the identification of individual words: When a string of four words was presented for 200 ms, followed by a cue to report the word in one of the four positions, this word was reported more accurately when the string was grammatical than when it was scrambled. They interpreted the SSE as reflecting parallel processing of the four words, rapid activation of a syntactic representation, and feedback from this syntactic representation to the word level. However, in a footnote Snell and Grainger (2017) reported a forced-choice version of the experiment in which grammaticality of the string did not have a significant effect on accuracy. By contrast, the classic Word Superiority Effect appears in forced choice, which rules out an account based on guessing. To further assess whether the SSE in free response is due to guessing, we conducted three on-line experiments (combined N = 629), varying the meaningfulness of the grammatical stimuli. Each participant completed both a free response block and a forced-choice block; in forced choice the target and foil words were similarly grammatical. We replicated an effect of the string's grammaticality in free response, but like Snell and Grainger (2017), did not find an effect in forced choice in any of our experiments. Thus, the SSE appears to be due to guessing based on linguistic knowledge, rather than parallel interactive activation.