Having considered the evidence that suggests a late locus of the effects of familiarity, we now turn to the contrasting evidence that suggests an earlier locus. Two examples will be considered in this regard. The first
Figure 5.5 Familiarity effects in word perception
Various ideas concerning where, in the processing system, familiarity exerts its influence (taken from Pollatsek et al., 1975).
Source: Pollatsek, A., Well, A. D., & Schindler, R. M. (1975).
Familiarity affects visual processing of words. Journal of
Experimental Psychology: Human Perception and Performance, 1,
328 –338 (fig. 1, p. 337). Reproduced with permission from APA.
‘What have we learnt?
The view that emerges from consideration of this body of work on familiarity effects in processing of letter strings is that any such effects reflect the operation, not of sensory encoding mechanisms, but of more central processes concerning storage, comparison and decision. More recently, though, other data have been collected and these suggest that not all effects of familiarity in word recogni-
tion are the same. Therefore there may indeed be grounds for arguing that sensory analysis can be influenced by the familiarity of the stimulus.
Remember, what we are trying to do here is ascertain what processing mechanisms are re- sponsible for particular behavioural effects – are familiarity effects (like the word frequency effect) indicative of perceptual or cognitive mechanisms?
pants actually did was to rate their confidence on a four-point scale as to how sure they were that the target occurred in interval X or that it occurred in interval Y. Part of the reason that this is such a clever experi- ment is that the participants were not being tested on their ability to identify the target letter strings; they were being tested on their ability to perceive where in the sequence the target appeared. In this regard, an important manipulation was that on half the trials the target letter string was a word and on half the trials the target string was a non-word. If participants had been tested on their ability to identify the letter strings then any difference in performance across familiar words and unfamiliar non-words may have been due to good guessing on the word trials. If the target is HOME and all you saw was HOM*, then why not guess HOME. In contrast, in their actual task Doyle and Leach (1988) measured perception of the letter strings in such a way that any influence of familiarity could only show up indirectly.
Indeed, the central result was that, generally speak- ing, judgements for words were more accurate than were judgements for non-words. Participants were more accurate in locating which interval a word occurred in than they were in locating the interval in which a non- word occurred in. There was a word superiority effect because participants were more accurate on word than non-word trials. More critical, though, was that the effect was revealed in measures relating to d′ and, as was explained above, this effect was therefore taken to reflect perceptual sensitivity and not response bias.
‘See ‘What have we learnt?’, page 155.
Merikle and Reingold (1990)
The second example of a ‘perceptual’ familiarity effect in word recognition is taken from a paper by Merikle
Figure 5.6 Schematic representation of the RSVP displays that Doyle and Leach (1988) used in their
experiments
The target occurred either before the pattern mask located at X or before the pattern mask located at Y. A target never appeared at both positions within a trial.
Source: Doyle, J. R., & Leach, C. (1988).
Word superiority in signal detection: Barely a glimpse, yet reading nonetheless. Cognitive Psychology, 20, 283–318 (fig. 1, p. 301). Reproduced with permission from Elsevier.
concerns something known as the word superiority
effect. As will become clear, it is perhaps misleading
to discuss the word superiority effect because there are many such effects that have been documented in the literature. All of these, however, share a common property, namely that performance with words is shown to be superior to performance with non-words across a range of different tasks. Here we take two studies and look at them in more detail.
Doyle and Leach (1988)
For instance, Doyle and Leach (1988) ran a study in which on trial a central fixation was initially presented prior to a rapid serial visual presentation (RSVP) sequence of display frames. Figure 5.6 provides a schematic of the various frames and timings used in the sequence. The two special frames were labelled as interval X and interval Y, respectively. X and Y were very brief, being in the order of 25 ms, and the target string of letters could appear in either interval on a given trial. As Figure 5.6 shows, immediately following each of these intervals was a rapid sequence of four masking displays. Collectively this sequence was known as a dynamic mask. In each frame of the dynamic mask, and at each target letter position, a different nonsense character was presented. Across the different frames of the masks different nonsense characters were presented and this created an impression of random characters rapidly changing in shape as the sequence unfolded.
All the participant had to do was to indicate whether the target letter string was presented in interval X or interval Y. Such a paradigm is known as a two-
alternative forced choice (a 2AFC) task. The partici-
pant is forced (okay, ‘encouraged’) to make a decision about one of two alternatives. In this case, what partici-
and Reingold (1990). They ran a series of experiments, only one of which will be mentioned here. On a given trial and following the offset of fixation display pre- sented to both eyes, a forward mask of a random string of letters was presented to the right eye for 50 ms. There was then a short pause followed by presentation of the target field to the left eye, which itself was followed by the mask being re-presented to the right eye. In the experiment under consideration the target frame duration was set at 50 ms. On half the trials, the target field contained a target letter string and on the remaining half trials the target field was blank.
For their first response on each trial participants merely had to report whether or not a target string was presented. This requirement was taken to define the
detection task. In addition, in a second lexical decision
task, participants then had to respond whether the tar-
get string was a word or a non-word. The experiment extends the basic methodology that Cheesman and Merikle (1984) developed, as we discussed in Chapter 4. Critically on the trials in which a target string was presented, a word appeared on a random half of the trials. On the remaining trials the target was a non- word. From the participant’s viewpoint, the experiment may have seemed a little odd because even on trials when they thought no target had been present they still were expected to make a lexical decision response! The data from the detection responses showed that detection performance was better for words than non-words; that is, there was again a robust word superiority effect. Participants were more accurate in deciding that a target string rather than a blank frame had been presented when the target string was a word than when it was a non-word. Again this result was
found when comparing scores computed as a variant of d′. Hence the conclusion was that stimulus famil- iarity did affect stimulus detection at the level of perceptual mechanisms. The findings could not be explained adequately in terms of a difference in bias in responding to words relative to non-words because of the effect being manifest through differences in d′ and not β. Word familiarity seems to exert an influence at a relatively early perceptual encoding stage of process- ing. Such evidence converges on the same conclusion as that put forward by Doyle and Leach (1988).
The findings on the lexical decision responses were equally provocative. Again measures of sensitivity were computed as a function of whether particip- ants were correct in their detection decisions. The data were broken down as to whether participants scored a hit or a miss on the detection task. Now it was found that participants generally performed well when they detected the target string (on hit trials). If they detected the target then they could judge reasonably well if it was a word or a non-word. A different pattern arose on trials where the participants missed the target. Now they were reasonably accurate in making a word response, but they were essentially guessing when the target was a non-word. If participants failed to detect the target, they were still reasonably good at responding that it might have been a word. Critically therefore, as Merikle and Reingold (1990) stated, ‘only familiar stimuli were perceived when participants failed to detect a stimulus but that both familiar and unfamiliar stimuli were perceived following stimulus detection’ (p. 582). The fact that participants could perform above chance with words that they were unable to detect was taken to show that some form of
‘What have we learnt?
In summary, what we have here is an effect of familiarity that shows that familiar items (words) are more easily detected than are unfamiliar items (non-words). This effect cannot be attributed to participants lowering some response criterion for words relative to non-words (better guessing on word trials) because the effects were revealed in d′ rather than b. The effect of word familiarity seems to be operating at the level of an early perceptual encoding stage of processing. We may conclude that at some perceptual level the processing of familiar items is facilitated relative to unfamiliar items.
In terms of real life, the suggestion is the follow- ing: assume you are standing on a railway platform and aimlessly watching the trains go by. What the experimental evidence suggests is that you are more likely to detect the presence of a friend or rel- ative on a train that moves past than someone you barely know. This is even though the presence of your friend is completely unexpected! The implica- tion of the work of Doyle and Leach (1988) is that such an effect actually does reflect something about perceptual encoding. Familiar items are more easily encoded than are unfamiliar items.
knowledge about words had been accessed without participants being consciously aware of this. Such an effect therefore is another indication of semantic acti- vation without conscious awareness – see Chapter 4.
Collectively, the data from the detection and lexical decision tasks reported by Merikle and Reingold (1990), has shown that stimulus familiarity affects both percep- tual and interpretative stages of processing. Familiar stimuli were detected better than unfamiliar stimuli. In addition, familiar stimuli, but not unfamiliar stimuli, were seen to engage with cognitive mechanisms con- cerned with stored knowledge. In this regard, there is a rather clear indication of how word familiarity operates: perceptual mechanisms are apparently tuned to the detection of familiar over unfamiliar stimuli. Furthermore, familiar stimuli apparently engage inter- pretive process in a way that unfamiliar stimuli do not.
‘See ‘What have we learnt?’, below.
from a briefly presented display than strings of random letters. More generally, where performance with words is better than with any other form of strings of letters.
word frequency effect Differences observed in word processing as a result of the frequency of occurrence in natural language. Common words are typically dealt with more effectively than rare words.
active theories A phrase used to refer to theories of perception in which the person can exert some control over how they perceive the world.
passive theories A phrase used to refer to theories of perception in which the person has no control over their perceptions of the world.
same/different judgement task Experimental paradigm where, on each trial, participants are required to compare two stimuli and respond as to whether they are the same or whether they differ from one another.
word superiority effect Generally speaking, when performance with words is superior relative to performance with non-words.
dynamic mask A mask that changes over time.
two-alternative forced choice Paradigm in which participants are required to choose between one of two possible alternatives on every trial.
detection task An experimental task in which the participant simply has to respond on every trial as to whether a stimulus is present or not.
Pinpoint question 5.4
If a word superiority effect shows up for the measure of d′ rather than b, does this mean perceptual or response mechanisms are responsible?
‘What have we learnt?
Initially the early work on familiarity effects was taken to show that familiarity operates at a level after sensory analysis is complete. In this regard the familiarity effects were interpreted as showing decision/response biases. Given this, it was con- cluded that familiarity exerts its influence at a post- perceptual stage of processing and therefore such effects tell us relatively little (perhaps nothing) about the nature of the perceptual system. The more recent studies, in contrast, have demonstrated quite convincingly that (at least where words are con- cerned) familiarity can exert an influence at an ear- lier perceptual stage of processing. Word familiarity conveys a benefit at the level of sensory analysis.
This is not to argue that the earlier and more recent data sets contradict one another. On the con- trary, what seems to have emerged is that certain tasks have revealed familiarity effects that do not reflect the operation of perceptual mechanisms, and as such they suggest that the corresponding fam- iliarity effects arise through the operation of post- perceptual mechanisms. Other tasks have quite clearly revealed how familiarity influences sensory analysis. On these grounds, it is most sensible to conclude that familiarity effects can arise at various stages throughout the processing system. Import- antly, there is evidence to suggest that sensory mechanisms may become particularly attuned to familiar stimuli.
mental category Stored knowledge regarding the nature of a class of objects or entities.
word recognition advantage For example, the finding that words are more accurately recognised
to be a critical factor in studies of expectancy and recency in perception.
In an early study Leeper (1935) initially presented participants with an unambiguous version of Boring’s figure and then went on to examine how the partici- pants would interpret a subsequent presentation of the corresponding ambiguous figure. A central result was that participants tended to report resolving the ambiguity in terms of the interpretation of the unam- biguous figure they had previously been given. So if they had been presented with the young woman version of the figure, they then interpreted the ambiguous figure as depicting a young woman. This was taken as evidence of how previous experience can affect present perceptions – participants tended to report the most recently accounted interpretation of the stimulus.