Edward Jayne
Edward Jayne
Style (vol. 18, No. 1 (Winter, 1984), pp. 83-97.
Slightly revised, August 1, 2004
The limitations of the short-term memory are of crucial importance in the experience of poetry. When reading poetry, we must invariably hold in consciousness those words which seem important to us, then forget them so the same can be done with subsequent words. Of course earlier words must be retained to give sufficient context to later ones, but it becomes impossible to focus our attention on these later words if too much of the earlier passage is kept in mind. The steady parade of new additions puts a relentless strain on our attention span, undermining our effort to fix in memory the residue of word combinations already experienced. As a result, there is an ongoing compromise as we shift our attention back and forth over the span of language to be remembered, then forgotten, in trying to keep up with poetry's forward momentum. The word-for-word sequence which splashes on consciousness is strictly dictated by syntagmatic relationships intended by the poet, with sequence just as important as syntactic relatedness, but this is trailed, and not very much later, by a second and more subjective inertia of word combinations which fade, dissipate, then become altogether sacrificed to the necessity of keeping abreast 1. Between these parallel waves of accretion and elimination there is a continually changing zone of awareness which constitutes our direct experience of poetry, mediating the interaction between the text itself and those resources of our memory and understanding which help us to experience its meaning.
To be emphasized is the difference in kind between the advance column and rear guard of word clusters as they march across one's mind almost completely out of step with each other. The linear accumulation of new words is totally dominated by the text since each word follows its predecessor in lock step exactly according to the poet's original intentions. However, the subsequent period of transition from attention to forgetting is much more likely to be affected by the reader's individual feelings, and the process of attrition can be expected to occur in sporadic and seemingly unpredictable combination 2. In fact, this transition would better fit the analogy of an army's desertion from the battlefield, with words disappearing sometimes one at a time, sometimes in thrown-together groups, even in whole platoons. All readers necessarily exercises their own preference of the residue they sustain in judging the implications of later words, but whatever positive choice they make also imposes negative consequences as to what must be forgotten. The lingering appreciation of a salient expression inevitably takes its toll in nearbyconstructions which are glossed over or perhaps altogether overlooked. And what rivets the attention of one reader might prove to be relatively insignificant for another. Since we all focus our attention, both in the choice and speed of what is forgotten becomes idiosyncratic. A different profile emerges not only between different readers, but also--if to a lesser extent--between different readings by the same readers. Nevertheless, good poetry probably succeeds with unusual effectiveness in structuring this entire process from beginning to end as words are received, integrated, and then forgotten in appropriate (if unrecognized) patterns of deterioration. In fact, the unique genius of such poets as Shakespeare, Keats, and Wallace Stevens can very likely be attributed to the wonderful resonance of their poetic diction--its contrived aftertaste (every bit as important as it is for good wine)--which takes place when new word combinations keep jostling into extinction the detritus of earlier ones.
Disregarded until now in stylistics, this function of attention overload is intrinsic to the experience of poetry and can be measured in both individual readers and groups of leaders, the latter through the use of averages. Short-term memory experiments can be devised to trace the pattern of memory loss, providing an entirely new approach to psychostylistics as a behavioral science which permits many concepts of literary criticism to be verified through the statistical interpretation of experimental data. Of course the text of poetry already supplies ample data of its own, but it comprises no more than the initial impact of words upon consciousness without providing any verifiable evidence of the process of abandonment which eliminates these words from our short-term memory. Ninety-nine out of a hundred readers who hear the six words "to be or not to be" can reformulate the sequence perfectly intact within the next few seconds. Professors who ask their students to write them down can expect to be rewarded with almost perfect unanimity, making statistics irrelevant to their inquiry. However, if Hamlet's soliloquy continues to be read so that new words incessantly spring into existence demanding their share of attention, it won't be long before a much smaller percentage of the class can remember these six words. At this point statistics becomes valid in comparing the relative short-term retention of words, groups of words, and poetic devices such as rhyme, rhythm, meter, and even tone and ambiguity. Exactly because old words must be forgotten to accommodate the new, there is sufficient decision making on the part of the reader to provide the basis for statistical quantification as opposed to the necessarily conjectural suppositions which have hitherto dominated stylistic analysis. A particular image or motif considered central to a poem's overall meaning, for example, might be established by short-term memory experiments to be so completely overshadowed by neighboring effects that it bears a very negligible role for most, if not all readers. When this happens, the critic's interpretive ingenuity--essentially a matter of alchemy--has been tested and abandoned through careful empirical investigation.
There are probably many ways to measure the short-term attention span in response to poetry, but the simplest and most productive seems to be the "constrained recall" technique of reading a passage, then suddenly stopping in order to list or paraphrase everything that can be remembered 3. With a class of students this might be accomplished on a group basis by a professor reading aloud up to a chosen cutoff point, then letting students transcribe everything they can remember from the quoted passage. If a larger sample is gathered from several classes, a tap recorder can be used to guarantee an identical oral recitation for all of the subjects used in the experiment 4. This is important in order to neutralize differences that might result from different intonation or a different emphasis for particular words in the separate readings.
My own research I want to summarize here is limited to response samples obtained from a single class of twenty-five freshman students immediately after hearing Shakespeare's Sonnets 98, 104 and 105, Yeats' "Lapis Lazuli," and a brief passage from Wordsworth's The Prelude, Book I, lines 425-33;
And in all the frosty season, when the sun
Was set, and visible for many a mile
The cottage windows blazed through twilight gloom.
I heeded not their summons: happy time
It was indeed for all of us--for me
It was a time of rapture! Clear and loud
The village clock tolled six,--I wheeled about,
Proud and exulting like an untired horse
That cares not for his home.
For Wordsworth's passage there were twenty-five response samples in which a predictable variety of jumbled transcriptions was obtained, as listed in Table 1, including protocol responses transcribed immediately after hearing the poem recited. Words incorrectly remembered are indicated by italics.
Table 1
1. clear and--the village clock struck six, I toiled, frost, in the dark frost
2. frosty windows twilight rapture The village clock struck six like an untired horse that cares not for his home.
3. summers evening enraptured horse that cares not for his home
4. And in the frosty summers eve the twilight filtered through the college window, and was quite caught in rapture which was synonymous with a tired horse returning home from the pasture.
5. one frosty evening when light came with the sun set even though one could see it light some through the cottage windows, everyone was happy. The clock struck six.
6. That cares not for its home wheeled about frosty the clock in the village square struck six.
7. And in the frosty evening I wandered home like a tired horse that cares not for his home. window.
8. horse cares not for his home summons frosty cottage windows rapture
9. horse cares not for his home summons frosty cottage windows rapture
10. cold winters night windows blazing I was feeling rapture town clock tolled six
11. cares, not, for, his, home, time, for, me, happy, time, rapture, I
12. cottage, lights, horse, home, unfurled, clock
13. raptured town like an untold horse who cares not for his
14. In the winter's frost . . . I cared little for home, like the untired horse who cares little for his home.
15. The village clock rang loud and clear. I wandered about like a lost horse searching for his home.
16. village, clock, tolled, six, wheeled, about, cottage. grove, I, though willows
17. on an autumn evening the sun (shone) across the field frosty windows on the little cottage I wheeled about like a proud and exulted master of the horse
18. I cared not for his home when it tolled six gloomy night I wheeled willow
19. Winter's day Sun reflects off of glazed windows. Cottage contains excitement within Can hear clock in the city. This man feels excitement.
20. like a un tired horse that cares not for his home. the bell struck 6 o'clock I was in rapture, invisible frost, winter, I cared not
21. And the clams Happy as we were and for me a long time of rapture the clock told 6 o'clock
22. time of rapture horse cares not for his home
23. proud and exulting like an uncared horse I headed home
24. Untired horse that cares not for his home glittering window city clock told 6 o'clock
25. untired horse that cares not for his home, frosty evening rapture, willows, windows
At first these transcriptions might seem hopelessly inadequate reconstructions of the original passage, therefore useless to the task of criticism. Nevertheless, they provide normal response samples, far more accurate than some of the specimens I have received from professional colleagues 5. Moreover, they provide useful evidence for documenting short-term memory overload once fragmentation and false reconstructions have been discounted as normal consequences of the forgetting process. Of course there is enough confusion to oblige extreme caution in making any generalizations, but the value of these transcriptions as experimental data cannot be underestimated once the variations have been sorted out. Statistics necessarily comes into play, and, with sufficient ingenuity on the part of the experimenter(s), many notions of literary criticism can actually be submitted to scientific verification.
Perhaps to their surprise, most subjects very quickly run out of the words and meaning they can remember, and when this happens they usually possess a clear sense that they have exhausted their memory. The quick sense of loss partly results from simple "decay," but its principal cause (from 85 to 90 percent, it is estimated), derives from "response competition" among words competing for their attention. Depending on the passage's overall length, written "protocol" samples can be expected to include from eight to twenty-five accurate words, seldom fewer than from five to six or more than from twenty to twenty-five. The longer the poem, the greater the number of words which can be remembered, but the unfavorable ratio between remembered and forgotten words significantly increases, and what is remembered seems more likely to depend upon concrete nouns (names, locations, etc.) which are accessible to one's long-term memory.
Also to be discounted is the often considerable disorder in protocol responses which can be traced to one or more of the following dislocations: 1) constructions which are recombined, for example an adjective borrowed from one noun to modify another, 2) words which are paraphrased, replaced by synonyms or even homonyms, 3) words which are used for different grammatical purpose, for example adjectives converted to nouns, and 4) words which have only an indirect connection with the poem through idiosyncratic associations triggered in the reader's mind. These four sources of confusion--and others as well--often lead to bizarre mixtures, but they may be discounted for statistical purposes by tabulating in their original positions all the words correctly indicated for the group as a whole in order to determine the relative salience (i.e., resistance to forgetting) of these words compared to nearby words in the same verbal context. Regardless of the inaccuracy and displacements which occur in all the particular samples, the extent to which words have endured in the memory of the group as a whole can be measured by the group's actual count that has been tabulated for these words. This is what lets us reconstruct the "average retention profile" of the passage once a sufficient number of samples has been gathered.
Response competition occurs when the transcription of certain words obliterates other words that could just as easily have been chosen for transcription. The concentration it takes to write down one word often drives from consciousness other words of comparable salience, so choice-making imposed by the experimental conditions compounds forgetfulness in the final selection of words. Mnemonic competition is unavoidably doubled, losses during the process of transcription automatically compounding those during the actual experience of the poetry. What results is a poetic equivalent to the Heisenberg Uncertainty Principle, since the documentation of evidence (i.e., its written transcription) obliges further distortion of this evidence. However, this additional burden of forgetting is useful to the task, since it augments choice making in protocol transcriptions. The words remembered are all the more likely to be those that were the most striking to the particular reader. This is true for each reader, but it bears even more significance for the group as a whole. The percentage of readers who retain in their short-term memories particular words and word combinations thus provides empirical evidence useful for statistical purposes in determining the relative salience of any word or group of words for most readers while reading or listening to the passage.
Tabulating procedures can be as rigid as seems necessary in recording protocol transcriptions so long as sufficient consistency is maintained in the application of these procedures. Mistakes can be overlooked in protocol responses if there are satisfactory guidelines for doing so on a regular basis. My own tabulations have been on the generous side, since I have treated each transcribed word as a "hit" rather than a "miss" if it resembles the original word in at least two of the following categories: a) meaning, b) sound, c) position, and d) grammatical use. Those words remembered with sufficient accuracy can then be tabulated in their original positions, permitting an average retention profile to be mapped out, whereupon the affective data ("affective" in the sense that response is emphasized rather than the text itself) can be confirmed on a statistical basis.
For the Wordsworth passage, for example, the twenty-five protocol samples listed above produce the "average" profile to be seen in Table 2:
Typically, this profile displays a "recency effect" (heavy emphasis on the latest words spoken) as well as a "primacy effect" (a lesser emphasis on the first words spoken) as well as an "asymptote" (or intermediate trough) that is located between the primacy and recency effects. The asymptote turns out to be the most important segment of the profile for highlighting those words and word combinations of the greatest salience among readers and listeners, since it singles out the words and word combinations whose attention survives the recency effect as well as the residual competition provided by the primacy effect. For my twenty-five students used as subjects, the words and word combinations with the greatest salience located in the asymptote included the cottage windows, twilight, happy time, time, rapture, and the village clock tolled six--I wheeled about. It is to be conceded that the immediate experience of this passage cannot be directly measured, but what is transcribed from memory can be tabulated. And then its tabulation can be submitted to statistical procedures in order to draw a variety of conclusions relevant to the experience itself.
As indicated earlier, what is perhaps the most noticeable about the protocols I have gathered is the stark inevitability of forgetting for even the briefest passage of poetry. For example, the Wordsworth passage consists of 69 words, yet it provides a mean of only 11.04 remembered words among the group tested, $16 percent of the total passage, with as standard deviation of 4.13. These results suggest that as many as 84% of readers can be expectedto remember fewer than sixteen words and as many as 98% of readers can be expected to remember fewer than nineteen words if word retention scores are distributed on a normal curve. Significantly, the three Shakespeare sonnets I have evaluated (98, 104, and 106) were from 60 to 70% longer in total word count, but they showed only a slight increase in the number of remembered words, while "Lapis Lazuli," by Yeats, five times as long, produced a mean of 21.7, less than twice as much. Its standard deviation of 6.20 indicates that 98% of readers in a normal distribution would remember fewer than thirty-five words, little over 10% of the poem. Moreover, its slightly improved number of retained words seems to have resulted from the long-term retention of 1) recognizable names, 2) "content words" which denote things and/or provide a sense of location, 3) heavily repetitive constructions such as "camel-back, horse-back, ass-back, mule-back, and 4) striking expressions such as "poets that are always gay," a line whose unintended homosexual implications dominated most of the protocols I received for the poem, very probably damaging its culminating impact intended by Yeats. Aside from these, the average word count of "Lapis Lazuli" dips again to approximately fifteen words, pretty much the same as for the other passages which have been tested.
One step further, successive readings can be undertaken with a sequence of cutoff points that trace the quick deflation of memory in response to poetry's oral recitation. This can be done by reading a passage several times to the same subjects without their having been exposed to the passage except in the sequence of experimental readings. Eventually this use of repetition can be expected to bias protocol responses, but a careful sequencing over the period of weeks (for example an entire semester), and with numerous other readings between, can minimize the impact of repetition for experimental purposes. As indicated by the frequency graph in Table 3, this is exactly what I have done with Wordsworth's passage, letting me superimpose three profiles with cutoff points at the words us, six, and home, sixteen words apart from each other. The results are as might be predicted:
heavy diminution can be seen to take place in the retention of particular words as the recency effect keeps pushing forward, but certain words obviously resist erosion with relative success. Between the second and third readings, for example for all of us was completely forgotten by everybody in the group, but for me dropped from 6 to 2, a 66% decline, while rapture dropped from 17 to 11, only a 33% decline. Among the words included in all three readings, heeded dropped from a low 4 to 1 followed by 0, whereas happy plummeted from an astronomical 22 in its position as the earliest word of the recency effect, to 6 followed by 3, suggesting a similar dropoff rate that finally settled into a lingering after-effect.
Unfortunately, the three-stage superimposition in Table 3 could only be constructed under difficult circumstances. The second reading (with the cutoff point at six, was taken two weeks after the first and included four earlier lines of poetry not indicated on the graph; the shortest reading (with the cutoff point at us was taken four weeks later, including an additional four lines of earlier poetry. My purpose in testing this sequence backwards was to minimize as much as possible the "repetition effect" linked with the recency effect, which would have been much more likely to have resulted in a growing familiarity with the passage. Nevertheless, after the latter two readings all my subjects claimed that they could vaguely remember their previous exposure to the passage. To the extent that they did, there might have been a slight increase in the accuracy of their transcriptions of earlier words, somewhat over-dramatizing the speed with which the passage was forgotten.
Obviously the recency effect dominates the short-term memory of oral recitation, as demonstrated by this particular passage from The Prelude in which the final ten words actually produced 45% of the total response items. For other passages the recency effect seems to have diminished in importance relative to both the length of the recitation and the heightened salience of earlier words. In Shakespeare's Sonnet 104, which is dominated by three earlier metaphors, the final ten words produced only 28% of the total response items. In "Lapis Lazuli," approximately three times as long as the sonnets, the recency effect was restricted to the final six words, and with these it only produced 12% of the total protocols. It can be speculated that the recency effect of longer passages somewhat loses its overall dominance but that its span is also likely to contract in length to such an extent that particular words, though fewer in number, might actually gain in relative salience. For an inverse correlation probably exists between the number of words sharing the recency effect and the relative impact of each of these words--the same inverse correlation which more or less occurs for the poem as a whole.
The recency effect seems sufficiently prolonged to contain at least a single line of pentameter verse, and often it extends to include immediately preceding words and phrases with close semantic and syntactic ties. The span of words composing the recency effect can be listed for the five passages used for this paper in order to illustrate its powerful impact in providing notional and syntactic coherence:
W Like an untired horse/That cares not for his home.
98 and, you away. /As with your shadow I with these did play.
104 age unbred: /Ere you born was beauty's summer dead.
106 Have eyes to wonder, but lack tongues to praise.
LL Their ancient, glittering eyes, are gay.
In two instances, "Lapis Lazuli" and Sonnet 106, the recency effect included the final line alone; in the other four instances the final line was supplemented by from two to seven earlier words. Though a sentence or clause was not always fully transcribed (for example in Sonnet 105 and the Wordsworth passages), the included grammatical construction nevertheless supported a fair degree of complexity.
Syntax seems to have been important in the overall pattern of what was remembered, but certain grammatical connections resisted forgetting more easily than others once the recency effect was no longer at work. In the Wordsworth passage, for example, the adjective-noun combinations frosty season, cottage windows, happy time, and village clock effectively resisted forgetting, as did the quick subject-predicate sequence, The village clock tolled six. The relative salience of adjective-noun combinations is illustrated by cottage windows, which is remembered as cottage windows, glittering windows, frosty windows, and frosty windows on the little cottage--in all but the first instance with words and images borrowed from elsewhere to fill in the missing attributive modifier. Verbs and adjectives in appositive constructions (for example proud and exulting) were more likely to be forgotten, and the only adjective which took precedence over its noun was frosty--possibly because it easily decomposed into the noun frost, as was indicated in several protocols, or because it could be associated with popular ice cream products. In this passage nouns had a 24% chance of being remembered, verbs and adjectives a 14% chance, and all other parts of speech roughly a 12% chance. The disproportionate advantage of remembered nouns (12.5% in excess of the number of nouns in the original passage) was about the same for the other poems I have tested, and I suspect it will be borne out in future experiments, probably because nouns are more likely to trigger personal associations independent of their syntactic function, for example with the word horse inspiring any number of memories relevant to one's experience with horses. If nothing else, this would suggest poets should be particularly careful with their use of nouns, avoiding both their excessive and deficient use relative to the words that contextualize them.
The quick decay of syntax subsequent to the recency effect can be indicated by comparing the rate at which the memory of word juxtapositions diminishes relative to the memory of particular words in the Wordsworth passage. The total number of correctly recalled juxtapositions can be divided by the total number of correctly recalled words to give what amounts to a syntax/word ratio. For this particular sample the ratio was approximately 90% for the half dozen words just before the cutoff point, but steadily dropped to less than 60% for the passage as a whole. If only 17% of the words were correctly remembered, the number of correct juxtapositions dropped even further to little over 10%, since many of the words were transcribed by subjects in almost a random sequence. What is suggested by these comparisons is that syntax was more likely to be forgotten than vocabulary in the recitation of the passage, leaving a jumble of stranded words whose sound and/or meaning augmented resistance to the process of forgetting.
Of the poetic devices impervious to memory loss, perhaps the most dramatic were images and metaphors, both of which gained considerable salience unless they were situated in a context which was unnecessarily confusing. The relative dominance of metaphors in Shakespeare's Sonnet 104, for instance, is indicated by their high item response total compared to the average for the sonnet as a whole:
159% Three winters cold Have from the forest shook three summer's pride
149% Three beauteous springs to yellow autumn turned
344% Three April perfumes in three hot Junes burned
The impact of these metaphors was apparently striking enough for them to have almost totally dominated the experience of the poem. However, figuration which was at all confusing seems to have been readily forgotten, and this tendency was compounded if there was strong competition from nearby images of greater clarity. For example, the three seasonal images listed above are followed by a fourth:
Ah yet doth beauty, like a dial-hand
Steal from his figure, and have no pace perceived.
The modern reader can only be confused by the lack of a clear relationship among beauty, dial-hand, steal, figure, and pace. What resulted for my group of students was a catastrophic drop in item response totals, only 11% of the average tally for the whole sonnet, and even less, only 2.5% of the April perfumes metaphor which immediately precedes it. It can be speculated that the April perfumes metaphor gains extra salience because the dial hand metaphor fails to draw much attention to itself. Once deprived of the recency effect, it quickly recedes from memory, and the might of conscious processing it might have received can be diverted elsewhere, in this instance retained by the metaphor immediately preceding it.
Rhyme also seems to have improved the resistance to forgetting in the poetry used for my experiments. For the three sonnets, 98, 104, and 106, the average retention of rhymed words compared with all others was 61% greater. This higher average was somewhat diminished by the inclusion of Sonnet 104, since it produces the same tally for rhymed words as for unrhymed words, probably because its rhyme scheme is overshadowed by its three central metaphors. In fact, the average response tally of its rhymed words was only 99% of that for its unrhymed words, as opposed to 231% and 214% increases respectively produced in sonnets 98 and 106. The striking effect of the metaphors in Sonnet 104 eclipsed all other poetic devices except for the euphuistic vowel repetition, eye I eyed, which evoked a whopping 320% improvement in retention.
Resistance to forgetting seems to have arisen from subjective factors as well as objective features of the poem itself. Apparently of major importance is the flexibility of the attention span in being fixed upon particular words for a moment or two, then rapidly brought forward to take into account later groups of words. Each reader is confronted with the ongoing decision what word combinations deserve close attention before darting ahead to perform the same operation with later ones. The more quickly one skips over a word, the more likely it will be forgotten; the longer one lets it linger in consciousness, the easier it will be recalled at a later time, but of course at the sacrifice of other words. Protocols dominated by the recency effect can accordingly be expected in two instances: from subjects who have just previously shifted their attention forward to the most recent words, and from subjects whose response to the passage has been uniformly carried forward commensurate to the rate at which its words of advanced upon each other. On the other hand, a better memory of earlier words can be expected from subjects temporarily caught up with whatever additional implications might have been suggested by them.
Perhaps to be expected, a clear bimodal distribution emerged among my students resulting from the choice they made between transcribing the most recent and earlier words. Among the twenty-five protocols for the Wordsworth passage, for example, eleven correctly indicated eight or more words from the last two lines, while nine transcribed absolutely nothing from these lines. Those emphasizing the recency effect produced an average of 13.82 correct words, almost twice as many as the 7.78% average for those who concentrated on the retention of earlier words. Moreover, the former were 80% accurate in their overall response as opposed to 60% for the latter. If long-term memory benefited from greater tenacity in the processing of earlier words, this was probably at the cost of both one's short-term retention of other words and the maximum possible residue of verbal experience to be retrieved at any particular cutoff point.
In other words, a tradeoff seems to have occurred. Savoring earlier word combinations detracted from what was later heard; on the other hand, sustained attention to the latest words discouraged better adequacy in weighing what had been heard before, but at the sacrifice of fully understanding what had been said before. As perhaps to be expected, a modest negative correlation seems to have existed between the short-term memory of poetry and verbal intelligence. With the Wordsworth passage, for example, students with verbal IQs measured at 110 or less furnished slightly more accurate protocols than those with verbal IQs of 120 or more. On the average, the 120-plus group produced 16.33 words as compared to 15.38 words for the 110-minus group, but fewer of these words were accurate for the 120-plus group, 11.44 to 11.50, producing an accuracy ratio of .72 compared to .81. For the overall group of twenty-five students there was a negative correlation of 1.24 between intelligence and total accurate words, probably because the 110-minus group more systematically restricted their transcriptions to the recency effect.
Subjective possibilities might be evaluated even more inclusively. Just as a large number of protocol transcriptions establishes a passage's unique profile as concentrated verbal experience, the single transcription replicates each reader's individual response pattern (or "signature") as characterized by his or her choice of material transcribed. Signification thus goes both ways: protocol transcriptions add up to define poetry's "average" conscious impact, but each protocol by itself also displays the stamp of personality brought to bear by each individual in processing its experience. It might take many hundreds of responses to determine a particular signature, but such becomes possible sooner or later.
Moreover, different averages might be obtained to contrast particular group responses to the same passage, for example when comparing men and women, old and young readers, right and left-handed readers, normal and schizophrenic readers, etc. To explore the possibility of establishing psychological correlations, I administered the MMPI (Minnesota Multiphasic Personality Inventory) to my students and went on what has been described as a fishing expedition to see if any obvious connections emerged. What I found, among other things, was a high correlation between schizophrenia (or Sc) scores and both the total number of accurate words they could recall from the Wordsworth passage and the total number of words dominated by its recency effect emphasizing an energetic horse unwilling to go home. These two correlations with were respectively .55 and .41 with reliability coefficients within the .01 and .05 range, and they were apparently confirmed by the .51 and .33 correlations for the MMPI F scale, measuring exaggerative tendencies often symptomatic of schizophrenic tendencies. None of the other passages produced comparable positive scores, and in fact Sonnet 106 received a .36 correlation on the Sc scale, so it seems possible to maintain on a very tentative basis that the Wordsworth passage, a narrative account of loneliness, somehow stirs the schizophrenic imagination with unusual resonance.
Variable response patterns might even be used to establish (or discount) profile "signatures" which differentiate particular groups, making it possible to use a sampling of protocol transcriptions to link an individual with one group or another, for example as a sociopath, paranoiac, etc., as is already done with TAT and Rorschach tests. For if poetry manifests a transaction between text and conscious response, protocol transcriptions based on short-term memory experiments necessarily integrate the two contexts--poetry and the reader's mind. An interface is documented that measures both. That repeated experimentation would be needed to isolate one from the other can be treated as a challenge, not as an insurmountable obstacle. Again, however, it should be stressed that much more experimentation would be necessary before any definitive generalization can be made.
It is to be conceded that research at this point is in its primitive formulations and can only be tentatively summarized. Additional research would very likely oblige numerous modifications in the rationale I have proposed here. Nevertheless, results have been plain enough in the few experiments I have conducted to suggest the possibility of testing a variety of theories of criticism by resorting to short term memory experiments. For example the "deviationist" theory of contextualized surprise--contexts and SD's (or stylistic devices)--proposed by such formalists as Victor Shklovsky, Max Eastman, Michael Riffaterre, and Stanley Fish at the beginning of his career can be refined by differentiating meaningful surprises from those which cause enough confusion to be forgotten all the more quickly. Similarly, it would be possible to confirm (or reject) my own theory of negative form proposed in "The Dialectics of Paranoid Form," Genre, 11 (Spring, 1978), pp. 131-57, based on concepts proposed in my two earlier College English articles, "Defense of the Homophobic Imagination," 37 (Sept., 1975), pp. 61-67, and "Up against the Mending Wall," 34 (April 1973), pp. 934-51. My suggestion that plot and theme often deny unacknowledged metaphoric implications might possibly be verified by short-term memory experiments if it turns out that words and phrases with a striking initial impact are completely forgotten, nevertheless suggesting complex attitudes that conflict with the easy solution offered by theme and plot. This would be a wonderful thing, whether or not my theory turns out to be valid. Critical theory, now almost a theology in its dependence on hieratic aesthetic pronouncements by received academic authorities, would actually become ever so slightly scientific--a source of data accessible to experimental confirmation.
For though the direct response to poetry might be too elusive to be captured in its full immediacy, the protocol transcriptions gathered as a secondary response can in fact be successfully measured. Despite their awkwardness, they afford concrete evidence generated by the reader's mind. Under certain obvious constraints, their interpretation can be of benefit to the craft of poetry, the study of this craft, and perhaps ultimately the improvement of the general public's reading skills in its response to this craft. If stylistic measurement has too often encouraged misleading distortions in the interpretation of poetry, a fully quantified statistical analysis of behavioral data relevant to the reader's response might actually restore the balance, confirming poetry's unique value on an entirely new basis in shaping and provoking meaningful experience in the reader's mind.
Footnotes
1. Only very few readers escape this imperative. Tales survive of Oscar Wilde's ability to scan a book as fast as he could turn its pages, then recite from memory any passage that was requested of him. Harold Bloom once demonstrated a comparable gift when he recited random pages of Hart Crane's volume of poetry backwards word-for-word despite the fact that had no consciously made any effort to memorize its contents backwards or forwards. Needless to say, both these individuals have dramatically surpassed the mnemonic limitation I am suggesting if accounts of the feats are true.
2. An ideal (and perhaps unrealizable) two-dimensional model of this reading process can be plotted on intersecting axes whose point of origin would be the silence preceding the first word. This word, then, would be located one unit above and to the right of the point of origin; the second word would be located two units above and to the right of it, and so on for all subsequent words of the poem. If the first word continues to be remembered when the second is spoken, its recollection would be represented by a "1" located directly above the original "1" which indicates its first utterance. If the word continues to be remembered when the third is spoken it would likewise be represented in the third row. The same goes for the second and all subsequent words. When a word is at last forgotten, it would still be represented in the higher rows, but by "0s" instead of "1s." The conscious experience of a poem would thus be diagrammed as a right triangle sitting on its apex, each word adding a new horizontal line of "1s" and "0s." The hypotenuse woudl represent the syntagmantic advancement of this poem, but the retention pattern at any cutoff point would be indicated by the horizontal pattern to the left of the spoken word. Meanwhile the lifespan in consciousness of any particular word would be indicated by the rising column of "Is" which terminates when "0s" are reached. Essentially the same paradigm may be compiled for groups of listeners by replacing "1s" and "0s" with total raw scores of all subjects, or, better yet, with percentages of those who can remember each old word as each new one gets added to the total.
3. Bennet B. Murdock, Human Memory: Theory and Data (Potomac, Maryland: Lawrence Eribaum Associates, 1974), p. 142. I obtain most of my nomenclature for memory theory from the Murdock text. I am also grateful to Jean Schwind and Fair Meeks for having helped tabulate several of the passages used for this paper, and to Professor Donald Ross for having helped to computerize my data in order to arrive at conclusions a good deal more sophisticated than would otherwise have been possible.
4. Tone and cadence can be expected to bear a significant impact upon the choice of words remembered in protocol transcriptions, so oral recitation should be smooth and without exaggeration or elongated pauses.
5. The emptiest protocols I have yet received were from a highly intelligent young philosophy professor at a major eastern university. Despite his best efort he could not transcribe more than three words apiece for the two passages I recited. His intense judgment of what he heard seems to have accelerated mnemonic loss to an unusual degree, especially in transcribing his words to the page.