The associability of CVC pairs.

A ssociability Norms page 1 Associability 1 (AS-1) 4 Associability 2 (AS-2) Associability 3 (AS-3) Results AS-1 and AS-2 Results AS-3 6 7 8 10 Discussion 13 Relation of AS to Performance . 14 Experiment 1 14 Experiment 2 18 Experiment 3 19 Experiment 4 21 Summary and Conclusions 24 References 25 Table 1 29 Table 2 30 Table 3 31 Table 4 63 Table 5 69 Table 6 70 Table 7 72 Table 8 73 Abstract In prior experiments subject-generated associative devices or natural language mediators (NLMs) linking pairs of items have been shown to faciliate acquisition of paired associates. Since Ss are questioned about NLMs after learning, such reports may be a result of the questioning.. Therefore, to obtain an a priori estimate of NLM probability this research was undertaken. Several hundred pairs, each composed of CVCs of about equal association value (AV), were shown for 15 sec. while Ss wrote down any NLM they could generate which linked both t.he stimulus and response. The AV level was varied between pairs. The proportion of Ss able to generate a NLM is the associability value (EIS). As expected, AS and AV are correlated although AS varies considerably among pairs composed of items about equal in AV. Experiments run after the AS scale was obtained demonstrated that AS is valuable as a predictor of learning rate. In addition, AS values were highly correlated with the frequency of NLMs in post-experiment reports. It was concluded that the AS measure represents a valuable addition to our understanding of the complexity of verbal learning.In prior experiments subject-generated associative devices or natural language mediators (NLMs) linking pairs of items have been shown to faciliate acquisition of paired associates. Since Ss are questioned about NLMs after learning, such reports may be a result of the questioning.. Therefore, to obtain an a priori estimate of NLM probability this research was undertaken. Several hundred pairs, each composed of CVCs of about equal association value (AV), were shown for 15 sec. while Ss wrote down any NLM they could generate which linked both t.he stimulus and response. The AV level was varied between pairs. The proportion of Ss able to generate a NLM is the associability value (EIS). As expected, AS and AV are correlated although AS varies considerably among pairs composed of items about equal in AV. Experiments run after the AS scale was obtained demonstrated that AS is valuable as a predictor of learning rate. In addition, AS values were highly correlated with the frequency of NLMs in post-experiment reports. It was concluded that the AS measure represents a valuable addition to our understanding of the complexity of verbal learning. -1011111111111111111111Millmikerrialiar........--__ The Associability of CVC Pairs' William E. Montague and Harold 0. Kiess2 University of Illinois, Urbana Associability Norms Meaningfulness of verbal material has been recognized as an important variable in paired-associate (PA) verbal learning and retention since Ebbinghaus' pioneering work in 1885. In fact, meaningfulness has been found to be the most powerful variable influencing speed of acquisition in verbal learning and has been the subject of a considerable amount of research in recent years. Simply, the effect of meaningfulness on learning rate can be seen in PA learning experiments where a pair like RIQKIV takes considerably longer to learn than a pair like TEL-COM. Since TEL and COM are items to which subjects (Ss) can give a relatively high number of associations we say they are more meaningful. Meaningfulness, or association value (AV), defined either in terms of the number of associations given in a limited time period, or the proportion of Ss giving an association to each verbal unit, has proved effective in influencing the rate of learning in many experiments (Goss and Nodine, 1965; Underwood and Schulz, 1960), Although there is no doubt that meaningfulness is a major variable, a number of results indicate that it cannot account for all the important phenomena. For example, in a number of experiments involved in the recent controversy over "one-trial" or "all-or-none" learning, the rate of learning for pairs of equivalent meaningfulness differed considerably (Underwood and Keppel, 1962). This difference between pairs cannot be Montague 2 based in any simple fashion upon meaningfulness since all pairs were originally equated on this dimension. In several recent experiments, during post-experiment questioning, Ss reported using various means such as sentences or words to associate or link items together. The ubiquity of such reports by our Ss, and by other experimenters' Ss (e.g., Bugelski, 1962; Clark, Lansford and Dallenbach, 1960; Runquist and Farley, 1964; Underwood, 1964, 1965; Underwood and Schulz, 1960) points to the need for systematic investigation of this mediation. From the reports it can be seen that Ss use various means of learning the pairs, and in addition, certain pairings are easier because Ss find them easier to associate. A single word may be used to link two items, or a phrase or sentence may be generated which includes the items, At other times Ss may learn a pair by recognizing that the items sound alike when pronounced or they may encode (transform) the items into words, We refer to these techniques as natural language mediators (NLMs) and have found that they are important in learning and recall (Adams and Montague, 1967; Kiess and Montague, 1965; Montague, Adams and Kiess, 1966; Montague and Wearing, 1967a, 1967b). This research has shown that S-generated associations between items, not the meaningfulness of individual items per se, produce superior retention in comparison to instances where such associations are not present. Unfortunately, our knowledgt. about NLMs is generally by means of an interview or questionnaire given upon completion of the experiment, asking Ss how they went about learning each pair. Subjective reports of this kind are suspect. It is possible that Ss construct answers to "please" the experimenter in accord with the demand characteristics of the experiment (Orne, 1962) so that the NLMs reported Montague 3 might not be accurate descriptions of what is learned. It is possible that some or all NLMs are not causal in learning, but may just be a correlate of the learning process (Adams, 1967). To ascertain their status in the associative process it is necessary that the probability of a NLM be manipulated independently of other variables known to affect learning rate. Therefore, some independent measure of the probability or likelihood of NLMs is necessary. Such a measure could be systematically manipulated to ascertain its relationship with other important variables and increase our understanding of verbal learning. Undoubtedly, items high in AV should be easier to link than those of low AV. Richardson and Erlebacher (1958) hypothesized that items high in AV have more associations and these associations can be used somehow to facilitate linking such pairs compared to pairs of lower AV. Their Ss' ratings of pairs generally agreed with the hypothesis. However, it is our contention that AV and ease of linking or associability, although correlated, are conceptually different and that this difference needs examination. Pairs of items of a given level of AV are liable to differ considerably in associability, and thereby, in ease of learning. To support these contentions we obtained quantitative estimates of the ease of associating various types of item pairs. We first attempted to determine the frequency with which Ss form associative connections between item pairs of different AV levels. The proportion of Ss generating a. NLM for each pair is the scale value of associability (AS). After the AS scale was obtained several experiments were undertaken to examine the effect of variation in AS on learning rate and retention. In addition, a subsequent scaling was undertaken to ascertain the role of stimulus and response AV in AS value, and to relate AS value to