Incidental Learning

The Effect of Levels of Processing

Upon Incidental Learning

W. Williams

Abstract

The aims of this experiment were to investigate the levels of processing model proposed by Craik and Lockhart (1972), and the study was based on their experiment. It was predicted that orthographic, phonemic and semantic word task would respectively result in higher levels of word recall. An opportunistic sample of 122 university students were randomly assigned to one of three conditions where simple word tasks required a different level of processing. The orthographic task required participants to count the vowels and syllables of stimulus words, the phonemic condition required rhyming words and the semantic condition required associated words. The same stimulus words were used for the three conditions, the independent variable being the instructions given. The participants were then subjected to an unexpected recall task of the stimulus words. Following a one way related ANOVA the results were found to be significant and the hypothesis was supported. The implications were discussed in the light of transfer appropriate processing and other factors possibly affecting the acquisition and retention of memory.

Introduction

During the sixties the multi-store structural approach dominated popular psychological preference as a viable system for the organisation of memory. The modal model proposed by Murdock (1967) and Atkinson and Schiffrin (1968, both cited in Craik and Lockhart 1972) described memory as dependant on the structure and limitations of the various 'stores'. The three main stores of sensory, short term and long term were distinguished through a specific form of coding, respectively visual, phonemic (Atkinson and Schiffrin 1968 as cited in Craik and Tulving 1972) and semantic, as well as retention span. A key feature of the process was the A to B to C format, in that anything within the long term store had previously been located within the sensory and short term stores. The sensory store was therefore responsible for incidental memory, acting as an initial buffer for all perceptual information, a preattentative perceptual register whereby stimuli can enter regardless of attention (Neisser 1967 as cited in Craik and Lockhart 1972).

Incidental learning as opposed to intentional learning whereby memory is consciously controlled or influenced offers an insight into the natural mechanics of preattentative memory without conscious control. It has therefore adopted a dominant focus within memory research, as a more objective view as to the factors affecting retention and recall is offered.

Although the multi-store approach was a seemingly effective paradigm for psychologists of the time, it soon began to develop criticism. Shallice and Warrington (1970 as cited in Craik and Lockhart 1972) suggested that the short term memory was not so much a through road as a diversion, with some information entering the long term memory immediately. The avoidance of the sensory store responsible for incidental memory creates a new variable responsible for the retention and coding of a memory. The retention of memory being a direct consequence of the perceptual processing involved evolved into an alternative approach known as levels of processing (Craik and Lockhart 1972). This offered an overlay to the multi-store approach. proposing that the retention and durability of memory is related to variable coding, although conversly there are no structural processes involved. Instead, the view is taken from a different angle whereby information can enter the sensory, STM or LTM directly, dependant on the level of process used as opposed to committed to it. It was suggested that seemingly shallow or superficial processing such as visual and auditory would result in a low retention span, whereas a deeper semantic level of processing would create a more durable and long term memory trace.

Craik and Lockhart devised a much replicated piece of research (1975) to illustrate their theory whereby participants were asked to perform a number of simple word tasks. The participants were split into three conditions, all of which were presented with the same sequence of unrelated words. The condition designed to process the words at a superficial level required the participants to answer yes or no as to whether the presented word was in upper or lower case. The condition to process the words auditorily required participants to answer yes or no as to whether a presented word rhymed with another word. The third condition were asked whether or not the presented word would fit into a given sentence. Half of the questions for all conditions were correct, and half incorrect (generating a NO answer). All of the participants were then asked to perform a memory test of the words used in the preceeding word tests, to the surprise of the unsuspecting participants. The results showed that semantically processed words were recalled significantly more than auditory words, which again were recalled significantly more than the visually processed words. This suggested that a deeper level of processing resulted in a more durable retention. However, the reaction times for answering the word questions also matched the retention levels. The visual tasks were very easy, whereas the semantic tasks required a little more thought and time. It could therefore be that the memory trace is more reliant on exposure and repetition than the level of processing used.

Craik and Lockhart tackled this problem in a further replication. This time, participants were given a more demanding task of visually processing words, whereby a consonant/vowel pattern was offered in a YES or NO situation. An example would be to match the word BOOK with the pattern CVVC. The semantic task was an easier version of the last one with simpler sentences, still yielding the same significant results and thus supporting the levels of processing theory.

A similar piece of research by Seaman and Murray (1976 as cited in Morris and Bransford 1977) required participants to either attend to a word's meaning, or attend to the position of the lips and tongue when speaking the stimulus word. The words processed semantically were recalled significantly more often than the more superficial processing. Also, when the meaningfulness of the words were manipulated across the conditions, the same results were found. This study shows that tasks of equal difficulty (if not the superficial processing more difficult) still create a significantly higher retention level for semantically processed words. However, as the tasks were so different in nature, it could be that the memory test lacked construct validity, as a test requiring the words simple recall was different in nature to the task of physical linguistics.

A replication of the original Craik and Lockhart (1975) experiment will be carried out, aiming to further test the validity and reliability of the theory of levels of processing. Participants will be randomly split into three conditions, processing a set of words orthographically, phonemically or semantically as the independent variable in a simple word task. The nature of experiment will again be kept from the participants until debriefing, the recall of the words will be the dependent variable.

The hypothesis states that more words will be recalled from the semantically processed words than the phonemically processed words, which again will be more significantly recalled than the orthographically processed words (one tailed).

The null hypothesis states that there will be no significant differences between the words recalled for the different levels of processing.

Method

Participants

The participants consisted of 122 first year undergraduate psychology students from Keele University. This opportunistic sample consisted of 21 males and 101 females. The age range of the participants was from 18 to 27 years. Participant counterbalancing was achieved through random assignment to the three experimental conditions, through random distribution of the three different questionnaires. This was seen as sufficient as the obvious participant variables were unlikely to affect recall.

Design

The design was an independent measures experiment. Although repeated measures would have been preferable to maintain a higher level of counterbalancing, this would have also endangered the secrecy as to the true nature of the experiment, as well as encourage possible practice, primacy or recency effects. The experimental conditions consisted of three separate groups of participants. The independent variable was the level of processing required to complete simple word tasks on question sheets, either orthographic, phonemic or semantic. The dependant variable was the amount of words recalled from the original list of words once the word tasks had been completed. The participant variables were counterbalanced through random assignment to the three independent variables. This was achieved through a sequenced order of questionnaires (first one orthographic, second one phonemic, third one semantic, fourth one orthographic etc.) before handing out to the participants. Each experimental condition consisted of 41 participants except for the orthographic condition which consisted of 40. The participants were all subjected to the same time scale for the tasks, 110 seconds. Other controls kept constant were the words for the levels of processing tasks which were the same for all participants regardless of condition. This was important to restrict biases towards remembering more elaborate, or shorter words for example. The time of day was the same for all subjects as tiredness would be a possible confounding variable affecting participant's retention span. The study was carried out early in the spring semester, at the start of a class which was relevant to the nature of the study.

Materials

Brief word questionnaires were used for the three conditions. Each participant in each condition received a sheet with twenty words, ten in each of two columns. After each word was an underlined space for the required task dependant upon the condition. At the top of the page were simple instructions, of which varied according to the condition. For the orthographic condition, participants were required to count the number of vowels and consonants. For the phonemic condition participants were to write a rhyming word of the stimulus word. The semantic group were to write a word somehow associated to the stimulus word, requiring the words meaning to be considered. Blank pages were also handed out for the second task of recall, on which the recalled words were to be written by the participants. Simple writing instruments were used by the participants to complete the tasks. An example question sheet is included in the appendix.

Procedure

The participants were informed that a simple word task was to be carried out, and that the task should be performed as quickly and accurately as possible. Once finished, the sheets were to be turned face down. No mention was given as to the true nature of the experiment. The question sheets were handed out to each of the participants, face down. Also, each participant received a blank sheet of paper. When instructed, each of the participants turned the sheets and began to complete the tasks. Once the time limit was reached (110 seconds) the participants were instructed to turn any exposed sheets face down again. Once all of the sheets were concealed, the participants were asked to turn their attention to the blank sheet. They were unpredictably instructed to recall and record as many words as possible from the original word list. Again, 110 seconds were given to complete the task. Once the time limit had exceeded, the participants were asked to count the number of words recalled, the amounts of which were then recorded by the experimenter for each experimental condition. The participants were informed as to the true nature of the experiment, in order to meet ethical guidelines set by the British Psychological Society.

Results

The raw data were collated by requesting the number of recalled words from the participants in each experimental condition. They are available from the appendix.

As only 40 participants were within the orthographic condition, an extra average word recall was added to equalise the number of participants for the other two conditions. This figure was six, and it was necessary for all conditions to be of equal number for the statistical analysis.

The average level of recall was 6.2 for the orthographic condition, 7.0 for the phonemic condition, and 9.7 for the semantic condition. These figures indicate that a higher amount of words ere recalled for the semantic processing condition than the phonemic condition. Also, a higher amount of words were recalled on average for the phonemic condition than the orthographic condition.

The variances between each condition were tested for significant differences using the one way analysis of variance. This was the most appropriate test for a one tailed experiment satisfying the parametric criteria, as three experimental conditions were used. The degrees of freedom were 40 for the between participant variable, and 2 for the between group variable. The level of probability chosen was 0.05, as it was decided this would be most effective at preventing type I and type II errors. As the hypothesis was one tailed the critical value of the F ratio was 19.47, for a P of 0.05.

F(2, 120) = 24.773 MSe = 139.6 P<0.05

Therefore the results were significant and the hypothesis was supported. The calculations for the ANOVA are in the appendix.

Discussion

The results supported the hypothesis. Therefore, significantly more words were recalled through semantically processing than phonemic processing, also through which significantly more words were recalled than through orthographic processing. These findings also support the work of Craik and Lockhart (1972), the study upon which this work was based. The results also support the work of similar studies such as Seaman and Murray (1976) where a more difficult task was offered for the phonemic task by concentrating on the position of the lips whilst speaking the stimulus words.

The study also supports another study which illustrates the levels of processing theory in a more applicable environment, that of Bower and Karlin (1974 as cited in Baddeley 1990). Participants were asked to judge the honesty of a person in a photograph. Another group were to judge the sex of the person within the photo. The task requiring judgement of honesty led to better recall in a recognition test. This suggests that the judgement is a more semantically based task where as sex recognition is a superficial task. However these results are open to interpretation. Both tasks can be seen as visual, as the 'meaning' or identity of the person within the photos are unknown to the participant. It is far more likely that the exposure to the images was a strong confounding variable. This study illustrates the power of levels of processing, but also highlights the possibility of other variables being involved.

The notion of exposure is related to rehearsal, which is thought to be an important variable influencing recall as well as elaboration. These concepts run parallel to the levels of processing, where rehearsal is seen as a superficial method to temporarily retain a memory trace which is then rapidly forgotten. Elaboration on the other hand is a far more durable method to imprint a memory trace. Craik and Lockhart (1975) agreed that levels of processing was not the only influencing factor upon long term memory. In one experiment they manipulated elaboration as the independent variable through simple sentences ('She cooked the.....') and complex ('The great bird swooped down and carried off the struggling.....'). For the sentences in which the stimulus word fitted the given sentence, the recall was twice as high for the complex condition compared to the simple. These results show that whilst both sentences required semantic processing but gave significantly different recall levels, there must be variables other than levels of processing that effect the memory trace, such as elaboration. Mechanic (1964 as cited in Baddeley 1990) researched the effect of rehearsal. Participants were given nonsense syllables as stimulus. One condition was required to pronounce the syllables only once, the other condition was required to repeat the syllables continuously within a given period of time. Again, the repeated condition recalled significantly more results than the other condition. Although only a superficial task, the retention was affected by another variable. However, the experiment lacked ecological validity as nonsense syllables are never usually encountered by the participant. On the other hand, words that are common could conversely affect the processing required. Overspill coding (Baddeley 1990) can lose the elegance of levels of processing experiments, by blurring the boundaries between the conditioned. For example finding a rhyme to the word dog will likely process the word semantically as opposed to strictly phonemically.

The effects of levels of processing are not only influenced by the input of a memory trace but also in the output, or recall. Morris Bransford and Franks (1977) replicated the work of Craik and Lockhart (1972). The recall however took place the following day. The results were the same as Craik and Lockhart's, semantic processing produced a significantly higher recall. However when given a stimulus word and asked if any words rhymed with it, participants recalled significantly more words from the phonemic condition. This research illustrates a concept known as transfer appropriate processing. It questions the validity of the recall tasks, as being more valid to the conditions hypothesised to create greater retention.

Another experiment to investigate transfer appropriate processing was carried out by Hyde and Jenkins (1969). One condition required subjects indicate whether a word could be fitted into a sentence. The other condition was required to find a rhyming word for a stimulus word. Half of the whole group of participants were given a recall test of the original words, in which the first, semantic condition produced higher recall. The other half were asked to recall words that rhymed with the stimulus words. This time, the group from the phonemic condition produced a higher word recall. Hyde and Jenkins concluded that the inferior memory trace was a result of the inappropriateness of the relationship between the acquisition and the recall test, rather than an inherent inferiority within the acquisition of memory traces, such as the perceived superficiality of a given incidental memory task.

In relation to this study, it appears that the implications of the results are open to a certain level of interpretation given the suggestions of past studies. The appropriateness of the memory task may act as a confounding variable, with the construct validity acting as a similar influence to the initial form of acquisition.

The close relation of the two confounding variables creates a more complex problem with regards to the definitions of either the acquisitional or retentional qualities. A circular definition is created as they both seem to be dependent on each other (Rummelhart and McClelland 1982 as cited in Baddeley 1990). The level of processing is defined by the retention it creates, even though the true level of processing is open to suggestion as it is difficult if not impossible to refine a task to a simple single level of processing. The retention level however is explained by the processing of the input, even though the retention is influenced by the memory task. This circular definition is a large problem with the levels of processing model, so much that Craik and Lockhart can not even find a solution.

Although questionable, the implications of the experiment are that a semantic level of processing will create a more durable memory trace. Simply visual input is a weak and unreliable method for recall. The problems with regards to the relationship between the word tasks and memory tasks does however offer applicational use, in that they are related. In order to remember something for a given situation, the acquisition should be in accordance with the recall. To illustrate in an ecologically valid example, consider the mechanics of learning for an examination. As the exam is likely to require understanding and intuition, it is more appropriate to revise for the exam by understanding the work and topics involved, as opposed to merely reading or rehearsing them. However if the exam is for a foreign language recall, it may be more appropriate to rehearse and speak the words, so that they will be recalled more naturally, rather than understanding the history of the origins of the foreign word. Another example is an implication of 'wanted' posters given on TV or newspapers for missing children or kidnappers. A visual image is a more appropriate method to be remembered by an individual rather than a psychological profile of the culprit.

These results and implications can be generalised to encompass a large selection of people and situations. The variables affecting acquisition and recall are proportionate within the individual, even if the abilities or experiences of the individual may differ.

The limitations of the study itself shouldn't have affected the results significantly as they support earlier studies, but should still be considered. Confounding variables that are difficult to control may have been an obstacle, such as the time required to complete the given tasks. Once established that other factors affect the acquisition of a word such as exposure (repetition) or elaboration, these should be controlled accordingly. The tasks were however quite carefully designed to be of equal difficulty. Other possible problems may lie with the type of participant involved. A faithful participant may have been tempted to cheat, by looking at a neighbours recalled words for cues of recognition, their own task sheet, or dishonesty when asked for the amount of recalled words. This problem may be overcome through individual testing although this would require more time and may be open to more confounding variables. Ethical considerations include the deceit of the participant with regards to the nature of the experiment, although this is essential to the incidental learning and is acceptable as long as a debriefing occurs after the experiment.

Other possible improvements may include offering more words and more time to process and recall. This will offer greater amounts to work with although may be vulnerable to primacy and recency effects (Glanzer and Cunitz 1966). Also, with greater time and money limitations, a greater number of subjects with inter condition counterbalancing would offer more representative and reliable results.

Various models for further improvements of the investigation arise from the problems encountered within the discussion. It would be interesting to offer various tests of recall, possibly replicating the work of Hyde and Jenkins (1969) where rhyming words were required to illustrate the effects of transfer appropriate processing. Designing a more difficult and time consuming task for the orthographic processing would clarify and offer greater weight to the results were they still significant for the recall of the semantically processed words. Also a study into the effects of rehearsal and elaboration would be interesting especially with regards to their relationship to levels of processing.

The investigation carried out produced significant results that support the hypothesis that semantic processing results in a more durable memory trace than more superficial processing. However, the results are open to interpretation in the light of previous studies. The transfer appropriate processing model researched by Hyde and Jenkins (1969) and Morris, Bransford and Franks (1977) suggests that the recall task is an equally affective variable as the acquisition task in recall. The two pole definition of processing then becomes circular (Rummelhart and McClelland 1982 as cited in Baddley 1990) as the retention level and the level of processing define each other. Also, other variables affecting the retention such as elaboration and rehearsal further blur the clarity of the findings. However these variables also refine the acquisition of memory traces for given situations which offers great implications for revision and learning.

References

Bernstein, Clarke-Stewart, Roy, Wickens. (1997) Psychology. Houghton Nifflin.

Gross, R.D. (1992) Psychology, the Science of Mind and Behaviour. Hodder and Stoughton.

Baddeley, A. (1990) Human Memory: Theory and Practice. Lawrence Erlbaum.

Wingfield & Byrnes (1987) The Psychology of Human Memory. Academic Press.

Morris, C.D, Bransford, J.D, & Franks, J.J, (1977), Levels of Processing versus transfer appropriate processing. Journal of Verbal Learning and Verbal Behaviour, 16, 519-533.

Craik, F.I.M, & Tulving, E. (1975) Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General, 104, 268-294.

Craik, F.I.M., & Lockhart, R.S. (1972) Levels of Processing: a framework for memory research. Journal of Verbal Learning and Verbal Behaviour, 11, 671-684