The Background

Jacoby and Whitehouse found that a brief masked prime can cause subjects to falsely identify "new" test words as having been seen. When subjects were aware of the prime's presentation the opposite effect occurred. They explained this phenomenon with the fluency heuristic, which is that when a stimulus is easily processed the individual feels familiar with the stimulus and thus may misidentify it as a stimulus that had been seen when in fact it had not.

Others, such as Bernstein and Welch (1991), state that these studies were flawed and that other factors may have resulted in the differences noted by Jacoby and Whitehouse. They argued that reported differences may be due to instructional differences, differences in the length of presentation of context words in the aware and unaware conditions, and because context word-test item stimulus onset asynchrony (SOA) for these conditions varied greatly in the duration of presentation.

So how do we establish that unconscious perception has occurred? Klinger describes two methods: Relative Sensitivity Strategy and Regression Strategy.

Relative Sensitivity Strategy (RSS) – compares direct and indirect measures

Direct – subjects judge whether a masked word or blank field was presented
Indirect – subjects make memory judgments and the effect of the masked priming on these memory judgments is measured.

According to the RSS approach, a finding of larger effects on the indirect measure than the direct measure would be strong evidence for an unconscious influence on the memory judgment.

Method

4 Phases:

• Eye dominance determination – using a double staircase technique word discrimination threshold was established for both eyes of each subject. Subjects were shown a word in one eye and a mask in the other, then asked to indicate which of four words were viewed via appropriate keystrokes. Correct responses resulted in a decrease in presentation time, while incorrect responses resulted in an increase in presentation time. Both staircases are run at the same time, in random order, so that the subject cannot anticipate the change in stimulus from trial to trial. The threshold is the average of 10 trials after the staircases have overlapped. This is done for each eye. The eye with the lower threshold is considered dominant and is presented with the mask throughout the experiment, while the other is presented with context words.
• Exposure phase – subjects were exposed to 90 of 180 test words for 1 second with a ½ second delay during which the screen appears blank.
• Indirect Measure or Memory Test phase – subjects were then presented with 180 words and asked to indicate which had been seen "old" and which were not seen "new," via appropriate keystrokes. Prior to the presentation of the test item, subjects were presented with a context word in the nondominant eye, while a mask or blank field is presented in the dominant eye, depending upon group assignment. Subjects were told that this was a warning and all subjects experienced 45 trials of each of the four possible presentation combinations: context identical and context unrelated with both old and new words.
• Direct Measure or Detection Test – masked condition subjects were then asked to discriminate between trials in which a word and mask were viewed and in which a mask and blank field were viewed. There were 100 trials of 50 words and 50 blank presentations. Subjects were then asked to indicate NO or YES via appropriate keystrokes, for whether or not a word was presented.

Computing Direct Measures of Perception – signal detection measure – d’, where the hit rate is the proportion of correct responses and the false alarm is the proportion of false responses.

Computing Indirect Measures of Perception – measure is a variant of d’, where a hit is responding "old" when an identical context word was presented and a false alarm is responding "old" when an unrelated context word was presented.

• Masked identical context words increase recognition judgment for "new" test items, but not "old" ones
• When not masked, context words do not significantly affect memory judgment for "old" or "new" items
• No significant difference between masked and nonmasked indirect effects for both "old" and "new" test words.

Relative Sensitivity Strategy – direct measure was significantly greater than the indirect measure for new and old test items (NO evidence that masked context words influence memory judgments)

Regression Strategy - intercepts were reliable for "new" test word trials, but not "old" test word trials, thus indicating evidence of unconscious perception. Although some error was noted, correctional procedures did not alter outcomes.

Supports Jacoby and Whitehouse’s (1989) assertion that masked words can increase false memory recognition, when regression strategy is used. This was not true for RSS.

Does not support Jacoby and Whitehouse’s (1989) results that clearly seen identical context words will decrease the false recognition of "new" words. Instead found similar effects for masked and nonmasked context words.

Method

Methods were the same except that attention to context words was manipulated through instruction to read aloud and remember context words, as was true for aware subjects in Jacoby and Whitehouse. This new subject group was identical to the nonmasked condition except for instruction. The population words were also changed to more closely match Jacoby and Whitehouse and counteract memory for previous test items.

Results

• Identical context words increased recognition judgments for "new" test items in the masked and the unattended condition.

• The attending condition made fewer "old" judgments for "new" test items preceded by identical context words.

Regression Strategy – the intercept was reliable for context words on "new" test word trials, but not "old" test word trials, which is seen as evidence of unconscious perception.

The differences observed in this experiment replicate those in Experiment 1 AND indicate that attentional differences may have resulted in the qualitative differences between aware and unaware groups that Jacoby and Whitehouse reported.

Study shows that nonmasked words can effect memory judgments like masked words.

Attending to or not attending to context words determines whether or not they increase or decrease false recognition.

Possible problems with the study:

• Having the detection trials after the memory test may have altered subjects’ thresholds.
• Retroactive priming may have occurred because test items primed the identical context words, thus making direct perception of masked words more likely.